From 3725cb1284f0c27fa1043474ebf4741f5510c715 Mon Sep 17 00:00:00 2001 From: ab Date: Thu, 15 Oct 2026 18:29:43 +0300 Subject: [PATCH 1/5] fix(instructions): repair four R3 contract defects and regenerate plugins MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Closes #357, Closes #355, Closes #354, Closes #356 One regeneration pass propagates four independent R3 source repairs to all 14 plugin targets (standard + lightweight): - #357: normalize the state-path token to coding-agents-prompting-flow-state.md and the step-7 producer/consumer identifier (Prompt Set, not Candidate Prompt Set) — resume and final validation handoff were broken in every distribution. - #355: replace the unconditional DeepWiki+Context7 mandate with a capability- aware source contract (prefer when available, explicit fallbacks otherwise); research README updated to match. - #354: remove nested subagent dispatch from the research skill, research-flow phase 3, and the research README — the researcher executes the approved prompt directly; parallel-slice needs are proposed back to the top-level orchestrator. - #356: define the small in-place modernization reduced route as an explicit, auditable exception to the phase-prerequisite contract (Phase 2 without Phase 1 only when no reference-library analysis is needed; Phase 6 without Phase 5 only under a bounded single-project final-review note). R2 (KTLO) carries equivalent text and is intentionally untouched — an R2 backport is a separate release-policy decision. --- instructions/r3/core/skills/research/README.md | 6 +++--- instructions/r3/core/skills/research/SKILL.md | 7 ++++--- .../r3/core/workflows/coding-agents-prompting-flow.md | 4 ++-- instructions/r3/core/workflows/modernization-flow.md | 2 +- instructions/r3/core/workflows/research-flow.md | 2 +- .../skills/coding-agents-prompting-flow/SKILL.md | 4 ++-- .../skills/modernization-flow/SKILL.md | 2 +- .../core-antigravity-light/skills/research-flow/SKILL.md | 2 +- plugins/core-antigravity-light/skills/research/README.md | 6 +++--- plugins/core-antigravity-light/skills/research/SKILL.md | 7 ++++--- .../skills/coding-agents-prompting-flow/SKILL.md | 4 ++-- .../core-antigravity/skills/modernization-flow/SKILL.md | 2 +- plugins/core-antigravity/skills/research-flow/SKILL.md | 2 +- plugins/core-antigravity/skills/research/README.md | 6 +++--- plugins/core-antigravity/skills/research/SKILL.md | 7 ++++--- plugins/core-claude-light/skills/research/README.md | 6 +++--- plugins/core-claude-light/skills/research/SKILL.md | 7 ++++--- .../workflows/coding-agents-prompting-flow.md | 4 ++-- plugins/core-claude-light/workflows/modernization-flow.md | 2 +- plugins/core-claude-light/workflows/research-flow.md | 2 +- plugins/core-claude/skills/research/README.md | 6 +++--- plugins/core-claude/skills/research/SKILL.md | 7 ++++--- .../core-claude/workflows/coding-agents-prompting-flow.md | 4 ++-- plugins/core-claude/workflows/modernization-flow.md | 2 +- plugins/core-claude/workflows/research-flow.md | 2 +- .../.agents/skills/coding-agents-prompting-flow/SKILL.md | 4 ++-- .../.agents/skills/modernization-flow/SKILL.md | 2 +- .../core-codex-light/.agents/skills/research-flow/SKILL.md | 2 +- plugins/core-codex-light/.agents/skills/research/README.md | 6 +++--- plugins/core-codex-light/.agents/skills/research/SKILL.md | 7 ++++--- .../.agents/skills/coding-agents-prompting-flow/SKILL.md | 4 ++-- .../core-codex/.agents/skills/modernization-flow/SKILL.md | 2 +- plugins/core-codex/.agents/skills/research-flow/SKILL.md | 2 +- plugins/core-codex/.agents/skills/research/README.md | 6 +++--- plugins/core-codex/.agents/skills/research/SKILL.md | 7 ++++--- .../commands/coding-agents-prompting-flow.md | 4 ++-- plugins/core-copilot-light/commands/modernization-flow.md | 2 +- plugins/core-copilot-light/commands/research-flow.md | 2 +- plugins/core-copilot-light/skills/research/README.md | 6 +++--- plugins/core-copilot-light/skills/research/SKILL.md | 7 ++++--- .../.github/prompts/coding-agents-prompting-flow.prompt.md | 4 ++-- .../.github/prompts/modernization-flow.prompt.md | 2 +- .../.github/prompts/research-flow.prompt.md | 2 +- .../.github/skills/research/README.md | 6 +++--- .../.github/skills/research/SKILL.md | 7 ++++--- .../.github/prompts/coding-agents-prompting-flow.prompt.md | 4 ++-- .../.github/prompts/modernization-flow.prompt.md | 2 +- .../.github/prompts/research-flow.prompt.md | 2 +- .../.github/skills/research/README.md | 6 +++--- .../.github/skills/research/SKILL.md | 7 ++++--- .../core-copilot/commands/coding-agents-prompting-flow.md | 4 ++-- plugins/core-copilot/commands/modernization-flow.md | 2 +- plugins/core-copilot/commands/research-flow.md | 2 +- plugins/core-copilot/skills/research/README.md | 6 +++--- plugins/core-copilot/skills/research/SKILL.md | 7 ++++--- .../commands/coding-agents-prompting-flow.md | 4 ++-- plugins/core-cursor-light/commands/modernization-flow.md | 2 +- plugins/core-cursor-light/commands/research-flow.md | 2 +- plugins/core-cursor-light/skills/research/README.md | 6 +++--- plugins/core-cursor-light/skills/research/SKILL.md | 7 ++++--- .../.cursor/commands/coding-agents-prompting-flow.md | 4 ++-- .../.cursor/commands/modernization-flow.md | 2 +- .../.cursor/commands/research-flow.md | 2 +- .../.cursor/skills/research/README.md | 6 +++--- .../.cursor/skills/research/SKILL.md | 7 ++++--- .../.cursor/commands/coding-agents-prompting-flow.md | 4 ++-- .../.cursor/commands/modernization-flow.md | 2 +- .../.cursor/commands/research-flow.md | 2 +- .../.cursor/skills/research/README.md | 6 +++--- .../.cursor/skills/research/SKILL.md | 7 ++++--- .../core-cursor/commands/coding-agents-prompting-flow.md | 4 ++-- plugins/core-cursor/commands/modernization-flow.md | 2 +- plugins/core-cursor/commands/research-flow.md | 2 +- plugins/core-cursor/skills/research/README.md | 6 +++--- plugins/core-cursor/skills/research/SKILL.md | 7 ++++--- 75 files changed, 165 insertions(+), 150 deletions(-) diff --git a/instructions/r3/core/skills/research/README.md b/instructions/r3/core/skills/research/README.md index a49556bfc..064fcd6b8 100644 --- a/instructions/r3/core/skills/research/README.md +++ b/instructions/r3/core/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/instructions/r3/core/skills/research/SKILL.md b/instructions/r3/core/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/instructions/r3/core/skills/research/SKILL.md +++ b/instructions/r3/core/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/instructions/r3/core/workflows/coding-agents-prompting-flow.md b/instructions/r3/core/workflows/coding-agents-prompting-flow.md index 0dbcbd0b3..dd77cb346 100644 --- a/instructions/r3/core/workflows/coding-agents-prompting-flow.md +++ b/instructions/r3/core/workflows/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/instructions/r3/core/workflows/modernization-flow.md b/instructions/r3/core/workflows/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/instructions/r3/core/workflows/modernization-flow.md +++ b/instructions/r3/core/workflows/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/instructions/r3/core/workflows/research-flow.md b/instructions/r3/core/workflows/research-flow.md index 6d1fe8c67..b7a681499 100644 --- a/instructions/r3/core/workflows/research-flow.md +++ b/instructions/r3/core/workflows/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-antigravity-light/skills/coding-agents-prompting-flow/SKILL.md b/plugins/core-antigravity-light/skills/coding-agents-prompting-flow/SKILL.md index bb878cfdd..d9ea16717 100644 --- a/plugins/core-antigravity-light/skills/coding-agents-prompting-flow/SKILL.md +++ b/plugins/core-antigravity-light/skills/coding-agents-prompting-flow/SKILL.md @@ -24,7 +24,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -86,7 +86,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md b/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md index d9fe500f2..0c43753b6 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md @@ -28,7 +28,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-antigravity-light/skills/research-flow/SKILL.md b/plugins/core-antigravity-light/skills/research-flow/SKILL.md index aad1fa0fd..2f6ffeee6 100644 --- a/plugins/core-antigravity-light/skills/research-flow/SKILL.md +++ b/plugins/core-antigravity-light/skills/research-flow/SKILL.md @@ -43,7 +43,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-antigravity-light/skills/research/README.md b/plugins/core-antigravity-light/skills/research/README.md index 9fb675c60..f75790ba3 100644 --- a/plugins/core-antigravity-light/skills/research/README.md +++ b/plugins/core-antigravity-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-antigravity-light/skills/research/SKILL.md b/plugins/core-antigravity-light/skills/research/SKILL.md index f5d079939..6bedde0fd 100644 --- a/plugins/core-antigravity-light/skills/research/SKILL.md +++ b/plugins/core-antigravity-light/skills/research/SKILL.md @@ -18,7 +18,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -41,13 +42,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-antigravity/skills/coding-agents-prompting-flow/SKILL.md b/plugins/core-antigravity/skills/coding-agents-prompting-flow/SKILL.md index bb878cfdd..d9ea16717 100644 --- a/plugins/core-antigravity/skills/coding-agents-prompting-flow/SKILL.md +++ b/plugins/core-antigravity/skills/coding-agents-prompting-flow/SKILL.md @@ -24,7 +24,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -86,7 +86,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-antigravity/skills/modernization-flow/SKILL.md b/plugins/core-antigravity/skills/modernization-flow/SKILL.md index d9fe500f2..0c43753b6 100644 --- a/plugins/core-antigravity/skills/modernization-flow/SKILL.md +++ b/plugins/core-antigravity/skills/modernization-flow/SKILL.md @@ -28,7 +28,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-antigravity/skills/research-flow/SKILL.md b/plugins/core-antigravity/skills/research-flow/SKILL.md index aad1fa0fd..2f6ffeee6 100644 --- a/plugins/core-antigravity/skills/research-flow/SKILL.md +++ b/plugins/core-antigravity/skills/research-flow/SKILL.md @@ -43,7 +43,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-antigravity/skills/research/README.md b/plugins/core-antigravity/skills/research/README.md index 9fb675c60..f75790ba3 100644 --- a/plugins/core-antigravity/skills/research/README.md +++ b/plugins/core-antigravity/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-antigravity/skills/research/SKILL.md b/plugins/core-antigravity/skills/research/SKILL.md index f5d079939..6bedde0fd 100644 --- a/plugins/core-antigravity/skills/research/SKILL.md +++ b/plugins/core-antigravity/skills/research/SKILL.md @@ -18,7 +18,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -41,13 +42,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-claude-light/skills/research/README.md b/plugins/core-claude-light/skills/research/README.md index a49556bfc..064fcd6b8 100644 --- a/plugins/core-claude-light/skills/research/README.md +++ b/plugins/core-claude-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-claude-light/skills/research/SKILL.md b/plugins/core-claude-light/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-claude-light/skills/research/SKILL.md +++ b/plugins/core-claude-light/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-claude-light/workflows/coding-agents-prompting-flow.md b/plugins/core-claude-light/workflows/coding-agents-prompting-flow.md index 556cc1aa8..c54d5c5fd 100644 --- a/plugins/core-claude-light/workflows/coding-agents-prompting-flow.md +++ b/plugins/core-claude-light/workflows/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-claude-light/workflows/modernization-flow.md b/plugins/core-claude-light/workflows/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-claude-light/workflows/modernization-flow.md +++ b/plugins/core-claude-light/workflows/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-claude-light/workflows/research-flow.md b/plugins/core-claude-light/workflows/research-flow.md index 03ceea8d4..2bc2b4816 100644 --- a/plugins/core-claude-light/workflows/research-flow.md +++ b/plugins/core-claude-light/workflows/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-claude/skills/research/README.md b/plugins/core-claude/skills/research/README.md index a49556bfc..064fcd6b8 100644 --- a/plugins/core-claude/skills/research/README.md +++ b/plugins/core-claude/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-claude/skills/research/SKILL.md b/plugins/core-claude/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-claude/skills/research/SKILL.md +++ b/plugins/core-claude/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-claude/workflows/coding-agents-prompting-flow.md b/plugins/core-claude/workflows/coding-agents-prompting-flow.md index 556cc1aa8..c54d5c5fd 100644 --- a/plugins/core-claude/workflows/coding-agents-prompting-flow.md +++ b/plugins/core-claude/workflows/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-claude/workflows/modernization-flow.md b/plugins/core-claude/workflows/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-claude/workflows/modernization-flow.md +++ b/plugins/core-claude/workflows/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-claude/workflows/research-flow.md b/plugins/core-claude/workflows/research-flow.md index 03ceea8d4..2bc2b4816 100644 --- a/plugins/core-claude/workflows/research-flow.md +++ b/plugins/core-claude/workflows/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-codex-light/.agents/skills/coding-agents-prompting-flow/SKILL.md b/plugins/core-codex-light/.agents/skills/coding-agents-prompting-flow/SKILL.md index 664732765..0ba957a23 100644 --- a/plugins/core-codex-light/.agents/skills/coding-agents-prompting-flow/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/coding-agents-prompting-flow/SKILL.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md b/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md index 74319010a..d6c1bd663 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-codex-light/.agents/skills/research-flow/SKILL.md b/plugins/core-codex-light/.agents/skills/research-flow/SKILL.md index 87a5ac82f..20db94499 100644 --- a/plugins/core-codex-light/.agents/skills/research-flow/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/research-flow/SKILL.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-codex-light/.agents/skills/research/README.md b/plugins/core-codex-light/.agents/skills/research/README.md index 9fb675c60..f75790ba3 100644 --- a/plugins/core-codex-light/.agents/skills/research/README.md +++ b/plugins/core-codex-light/.agents/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-codex-light/.agents/skills/research/SKILL.md b/plugins/core-codex-light/.agents/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-codex-light/.agents/skills/research/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-codex/.agents/skills/coding-agents-prompting-flow/SKILL.md b/plugins/core-codex/.agents/skills/coding-agents-prompting-flow/SKILL.md index 664732765..0ba957a23 100644 --- a/plugins/core-codex/.agents/skills/coding-agents-prompting-flow/SKILL.md +++ b/plugins/core-codex/.agents/skills/coding-agents-prompting-flow/SKILL.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md b/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md index 74319010a..d6c1bd663 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-codex/.agents/skills/research-flow/SKILL.md b/plugins/core-codex/.agents/skills/research-flow/SKILL.md index 87a5ac82f..20db94499 100644 --- a/plugins/core-codex/.agents/skills/research-flow/SKILL.md +++ b/plugins/core-codex/.agents/skills/research-flow/SKILL.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-codex/.agents/skills/research/README.md b/plugins/core-codex/.agents/skills/research/README.md index 9fb675c60..f75790ba3 100644 --- a/plugins/core-codex/.agents/skills/research/README.md +++ b/plugins/core-codex/.agents/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-codex/.agents/skills/research/SKILL.md b/plugins/core-codex/.agents/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-codex/.agents/skills/research/SKILL.md +++ b/plugins/core-codex/.agents/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-copilot-light/commands/coding-agents-prompting-flow.md b/plugins/core-copilot-light/commands/coding-agents-prompting-flow.md index 2900faac7..03f937315 100644 --- a/plugins/core-copilot-light/commands/coding-agents-prompting-flow.md +++ b/plugins/core-copilot-light/commands/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-copilot-light/commands/modernization-flow.md b/plugins/core-copilot-light/commands/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-copilot-light/commands/modernization-flow.md +++ b/plugins/core-copilot-light/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-copilot-light/commands/research-flow.md b/plugins/core-copilot-light/commands/research-flow.md index 4d07fc32d..9776c6d6d 100644 --- a/plugins/core-copilot-light/commands/research-flow.md +++ b/plugins/core-copilot-light/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-copilot-light/skills/research/README.md b/plugins/core-copilot-light/skills/research/README.md index 2bc716ee7..c8a1da6f2 100644 --- a/plugins/core-copilot-light/skills/research/README.md +++ b/plugins/core-copilot-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-copilot-light/skills/research/SKILL.md b/plugins/core-copilot-light/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-copilot-light/skills/research/SKILL.md +++ b/plugins/core-copilot-light/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-copilot-standalone-light/.github/prompts/coding-agents-prompting-flow.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/coding-agents-prompting-flow.prompt.md index 2900faac7..03f937315 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/coding-agents-prompting-flow.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/coding-agents-prompting-flow.prompt.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-copilot-standalone-light/.github/prompts/research-flow.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/research-flow.prompt.md index 4d07fc32d..9776c6d6d 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/research-flow.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/research-flow.prompt.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-copilot-standalone-light/.github/skills/research/README.md b/plugins/core-copilot-standalone-light/.github/skills/research/README.md index 5972aa05f..ac8235e83 100644 --- a/plugins/core-copilot-standalone-light/.github/skills/research/README.md +++ b/plugins/core-copilot-standalone-light/.github/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `prompts/self-help-flow.prompt.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md b/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md +++ b/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-copilot-standalone/.github/prompts/coding-agents-prompting-flow.prompt.md b/plugins/core-copilot-standalone/.github/prompts/coding-agents-prompting-flow.prompt.md index 2900faac7..03f937315 100644 --- a/plugins/core-copilot-standalone/.github/prompts/coding-agents-prompting-flow.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/coding-agents-prompting-flow.prompt.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-copilot-standalone/.github/prompts/research-flow.prompt.md b/plugins/core-copilot-standalone/.github/prompts/research-flow.prompt.md index 4d07fc32d..9776c6d6d 100644 --- a/plugins/core-copilot-standalone/.github/prompts/research-flow.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/research-flow.prompt.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-copilot-standalone/.github/skills/research/README.md b/plugins/core-copilot-standalone/.github/skills/research/README.md index 5972aa05f..ac8235e83 100644 --- a/plugins/core-copilot-standalone/.github/skills/research/README.md +++ b/plugins/core-copilot-standalone/.github/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `prompts/self-help-flow.prompt.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-copilot-standalone/.github/skills/research/SKILL.md b/plugins/core-copilot-standalone/.github/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-copilot-standalone/.github/skills/research/SKILL.md +++ b/plugins/core-copilot-standalone/.github/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-copilot/commands/coding-agents-prompting-flow.md b/plugins/core-copilot/commands/coding-agents-prompting-flow.md index 2900faac7..03f937315 100644 --- a/plugins/core-copilot/commands/coding-agents-prompting-flow.md +++ b/plugins/core-copilot/commands/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-copilot/commands/modernization-flow.md b/plugins/core-copilot/commands/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-copilot/commands/modernization-flow.md +++ b/plugins/core-copilot/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-copilot/commands/research-flow.md b/plugins/core-copilot/commands/research-flow.md index 4d07fc32d..9776c6d6d 100644 --- a/plugins/core-copilot/commands/research-flow.md +++ b/plugins/core-copilot/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-copilot/skills/research/README.md b/plugins/core-copilot/skills/research/README.md index 2bc716ee7..c8a1da6f2 100644 --- a/plugins/core-copilot/skills/research/README.md +++ b/plugins/core-copilot/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-copilot/skills/research/SKILL.md b/plugins/core-copilot/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-copilot/skills/research/SKILL.md +++ b/plugins/core-copilot/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-cursor-light/commands/coding-agents-prompting-flow.md b/plugins/core-cursor-light/commands/coding-agents-prompting-flow.md index 5992b1ed3..21418c4af 100644 --- a/plugins/core-cursor-light/commands/coding-agents-prompting-flow.md +++ b/plugins/core-cursor-light/commands/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-cursor-light/commands/modernization-flow.md b/plugins/core-cursor-light/commands/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-cursor-light/commands/modernization-flow.md +++ b/plugins/core-cursor-light/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-cursor-light/commands/research-flow.md b/plugins/core-cursor-light/commands/research-flow.md index bdace9f9b..8f0f040cf 100644 --- a/plugins/core-cursor-light/commands/research-flow.md +++ b/plugins/core-cursor-light/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-cursor-light/skills/research/README.md b/plugins/core-cursor-light/skills/research/README.md index a8e900058..fb36adf58 100644 --- a/plugins/core-cursor-light/skills/research/README.md +++ b/plugins/core-cursor-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-cursor-light/skills/research/SKILL.md b/plugins/core-cursor-light/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-cursor-light/skills/research/SKILL.md +++ b/plugins/core-cursor-light/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/coding-agents-prompting-flow.md b/plugins/core-cursor-standalone-light/.cursor/commands/coding-agents-prompting-flow.md index 5992b1ed3..21418c4af 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/coding-agents-prompting-flow.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/research-flow.md b/plugins/core-cursor-standalone-light/.cursor/commands/research-flow.md index bdace9f9b..8f0f040cf 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/research-flow.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md b/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md index a8e900058..fb36adf58 100644 --- a/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md +++ b/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md b/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md +++ b/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-cursor-standalone/.cursor/commands/coding-agents-prompting-flow.md b/plugins/core-cursor-standalone/.cursor/commands/coding-agents-prompting-flow.md index 5992b1ed3..21418c4af 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/coding-agents-prompting-flow.md +++ b/plugins/core-cursor-standalone/.cursor/commands/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-cursor-standalone/.cursor/commands/research-flow.md b/plugins/core-cursor-standalone/.cursor/commands/research-flow.md index bdace9f9b..8f0f040cf 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/research-flow.md +++ b/plugins/core-cursor-standalone/.cursor/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-cursor-standalone/.cursor/skills/research/README.md b/plugins/core-cursor-standalone/.cursor/skills/research/README.md index a8e900058..fb36adf58 100644 --- a/plugins/core-cursor-standalone/.cursor/skills/research/README.md +++ b/plugins/core-cursor-standalone/.cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md b/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md +++ b/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-cursor/commands/coding-agents-prompting-flow.md b/plugins/core-cursor/commands/coding-agents-prompting-flow.md index 5992b1ed3..21418c4af 100644 --- a/plugins/core-cursor/commands/coding-agents-prompting-flow.md +++ b/plugins/core-cursor/commands/coding-agents-prompting-flow.md @@ -26,7 +26,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 5. Orchestrator and subagents MUST USE SKILL `coding-agents-prompt-authoring`. 6. No rush, take your time, MUST FOLLOW WORKFLOW ENTIRELY, no skipping. 7. This workflow MUST be used with Fable, Opus, GPT-6.1-Sol+ class models => IF NOT - DEMAND USER TO SWITCH MODEL. -8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state` file. +8. Workflow state MUST be saved to `agents/TEMP//coding-agents-prompting-flow-state.md` file. @@ -88,7 +88,7 @@ Orchestrates prompt authoring/adaptation via `discover -> extract+intake -> blue 1. Validate final artifacts against intent, contracts, failure modes, and traceability. -2. Input: `Prompt Brief` + `Blueprint` + `Candidate Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). +2. Input: `Prompt Brief` + `Blueprint` + `Prompt Set` + `Simulation Notes`. Output: `Final Prompt Set` + `Validation Pack` (`Checklist Results`, `Tests`, `Failure Modes`, `Traceability`; persistent report optional in FEATURE PLAN folder as validation-report.md). 3. Update `coding-agents-prompting-flow-state.md` and mark complete. 4. HITL final approval required before persistence; small tasks may stay in-memory and be returned in message. diff --git a/plugins/core-cursor/commands/modernization-flow.md b/plugins/core-cursor/commands/modernization-flow.md index 4a9edaa8b..621041514 100644 --- a/plugins/core-cursor/commands/modernization-flow.md +++ b/plugins/core-cursor/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview diff --git a/plugins/core-cursor/commands/research-flow.md b/plugins/core-cursor/commands/research-flow.md index bdace9f9b..8f0f040cf 100644 --- a/plugins/core-cursor/commands/research-flow.md +++ b/plugins/core-cursor/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt as a separate subagent. +1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-cursor/skills/research/README.md b/plugins/core-cursor/skills/research/README.md index a8e900058..fb36adf58 100644 --- a/plugins/core-cursor/skills/research/README.md +++ b/plugins/core-cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. +Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use DeepWiki and Context7" — a named-tool mandate written into the generated research prompt, not a suggestion. +- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-cursor/skills/research/SKILL.md b/plugins/core-cursor/skills/research/SKILL.md index eb7a515ea..2941a0798 100644 --- a/plugins/core-cursor/skills/research/SKILL.md +++ b/plugins/core-cursor/skills/research/SKILL.md @@ -21,7 +21,8 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +45,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- Spawn parallel subagents to go over individual ideas or areas +- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use DeepWiki and Context7 +2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) From 2c67ab10b50248a4eaaabcc96d8f12ec5e481ae9 Mon Sep 17 00:00:00 2001 From: ab Date: Thu, 15 Oct 2026 18:44:04 +0300 Subject: [PATCH 2/5] fix(instructions): scope research execution rule to the caller (CI: Validate Prompts) The prompt-quality auditor flagged a High conflict: the unconditional 'execute directly, never spawn subagents' rule broke the top-level team-manager caller, which activates research directly and must keep dispatching the prepared prompt to a researcher subagent. Rule is now executor-scoped: spawned researcher executes locally and proposes parallel slices to its parent; top-level agent dispatches under orchestration. Nested dispatch stays prohibited. Regenerated across all 14 plugin targets. --- instructions/r3/core/skills/research/README.md | 2 +- instructions/r3/core/skills/research/SKILL.md | 4 ++-- plugins/core-antigravity-light/skills/research/README.md | 2 +- plugins/core-antigravity-light/skills/research/SKILL.md | 4 ++-- plugins/core-antigravity/skills/research/README.md | 2 +- plugins/core-antigravity/skills/research/SKILL.md | 4 ++-- plugins/core-claude-light/skills/research/README.md | 2 +- plugins/core-claude-light/skills/research/SKILL.md | 4 ++-- plugins/core-claude/skills/research/README.md | 2 +- plugins/core-claude/skills/research/SKILL.md | 4 ++-- plugins/core-codex-light/.agents/skills/research/README.md | 2 +- plugins/core-codex-light/.agents/skills/research/SKILL.md | 4 ++-- plugins/core-codex/.agents/skills/research/README.md | 2 +- plugins/core-codex/.agents/skills/research/SKILL.md | 4 ++-- plugins/core-copilot-light/skills/research/README.md | 2 +- plugins/core-copilot-light/skills/research/SKILL.md | 4 ++-- .../.github/skills/research/README.md | 2 +- .../.github/skills/research/SKILL.md | 4 ++-- .../core-copilot-standalone/.github/skills/research/README.md | 2 +- .../core-copilot-standalone/.github/skills/research/SKILL.md | 4 ++-- plugins/core-copilot/skills/research/README.md | 2 +- plugins/core-copilot/skills/research/SKILL.md | 4 ++-- plugins/core-cursor-light/skills/research/README.md | 2 +- plugins/core-cursor-light/skills/research/SKILL.md | 4 ++-- .../.cursor/skills/research/README.md | 2 +- .../.cursor/skills/research/SKILL.md | 4 ++-- .../core-cursor-standalone/.cursor/skills/research/README.md | 2 +- .../core-cursor-standalone/.cursor/skills/research/SKILL.md | 4 ++-- plugins/core-cursor/skills/research/README.md | 2 +- plugins/core-cursor/skills/research/SKILL.md | 4 ++-- 30 files changed, 45 insertions(+), 45 deletions(-) diff --git a/instructions/r3/core/skills/research/README.md b/instructions/r3/core/skills/research/README.md index 064fcd6b8..9a3fe7493 100644 --- a/instructions/r3/core/skills/research/README.md +++ b/instructions/r3/core/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/instructions/r3/core/skills/research/SKILL.md b/instructions/r3/core/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/instructions/r3/core/skills/research/SKILL.md +++ b/instructions/r3/core/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-antigravity-light/skills/research/README.md b/plugins/core-antigravity-light/skills/research/README.md index f75790ba3..79b9cced5 100644 --- a/plugins/core-antigravity-light/skills/research/README.md +++ b/plugins/core-antigravity-light/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-antigravity-light/skills/research/SKILL.md b/plugins/core-antigravity-light/skills/research/SKILL.md index 6bedde0fd..a90cc66b1 100644 --- a/plugins/core-antigravity-light/skills/research/SKILL.md +++ b/plugins/core-antigravity-light/skills/research/SKILL.md @@ -18,7 +18,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -42,7 +42,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-antigravity/skills/research/README.md b/plugins/core-antigravity/skills/research/README.md index f75790ba3..79b9cced5 100644 --- a/plugins/core-antigravity/skills/research/README.md +++ b/plugins/core-antigravity/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-antigravity/skills/research/SKILL.md b/plugins/core-antigravity/skills/research/SKILL.md index 6bedde0fd..a90cc66b1 100644 --- a/plugins/core-antigravity/skills/research/SKILL.md +++ b/plugins/core-antigravity/skills/research/SKILL.md @@ -18,7 +18,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -42,7 +42,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-claude-light/skills/research/README.md b/plugins/core-claude-light/skills/research/README.md index 064fcd6b8..9a3fe7493 100644 --- a/plugins/core-claude-light/skills/research/README.md +++ b/plugins/core-claude-light/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-claude-light/skills/research/SKILL.md b/plugins/core-claude-light/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-claude-light/skills/research/SKILL.md +++ b/plugins/core-claude-light/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-claude/skills/research/README.md b/plugins/core-claude/skills/research/README.md index 064fcd6b8..9a3fe7493 100644 --- a/plugins/core-claude/skills/research/README.md +++ b/plugins/core-claude/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-claude/skills/research/SKILL.md b/plugins/core-claude/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-claude/skills/research/SKILL.md +++ b/plugins/core-claude/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-codex-light/.agents/skills/research/README.md b/plugins/core-codex-light/.agents/skills/research/README.md index f75790ba3..79b9cced5 100644 --- a/plugins/core-codex-light/.agents/skills/research/README.md +++ b/plugins/core-codex-light/.agents/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-codex-light/.agents/skills/research/SKILL.md b/plugins/core-codex-light/.agents/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-codex-light/.agents/skills/research/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-codex/.agents/skills/research/README.md b/plugins/core-codex/.agents/skills/research/README.md index f75790ba3..79b9cced5 100644 --- a/plugins/core-codex/.agents/skills/research/README.md +++ b/plugins/core-codex/.agents/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-codex/.agents/skills/research/SKILL.md b/plugins/core-codex/.agents/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-codex/.agents/skills/research/SKILL.md +++ b/plugins/core-codex/.agents/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-copilot-light/skills/research/README.md b/plugins/core-copilot-light/skills/research/README.md index c8a1da6f2..4b2131ae3 100644 --- a/plugins/core-copilot-light/skills/research/README.md +++ b/plugins/core-copilot-light/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-copilot-light/skills/research/SKILL.md b/plugins/core-copilot-light/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-copilot-light/skills/research/SKILL.md +++ b/plugins/core-copilot-light/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-copilot-standalone-light/.github/skills/research/README.md b/plugins/core-copilot-standalone-light/.github/skills/research/README.md index ac8235e83..f7b492df5 100644 --- a/plugins/core-copilot-standalone-light/.github/skills/research/README.md +++ b/plugins/core-copilot-standalone-light/.github/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `prompts/self-help-flow.prompt.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md b/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md +++ b/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-copilot-standalone/.github/skills/research/README.md b/plugins/core-copilot-standalone/.github/skills/research/README.md index ac8235e83..f7b492df5 100644 --- a/plugins/core-copilot-standalone/.github/skills/research/README.md +++ b/plugins/core-copilot-standalone/.github/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `prompts/self-help-flow.prompt.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-copilot-standalone/.github/skills/research/SKILL.md b/plugins/core-copilot-standalone/.github/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-copilot-standalone/.github/skills/research/SKILL.md +++ b/plugins/core-copilot-standalone/.github/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-copilot/skills/research/README.md b/plugins/core-copilot/skills/research/README.md index c8a1da6f2..4b2131ae3 100644 --- a/plugins/core-copilot/skills/research/README.md +++ b/plugins/core-copilot/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-copilot/skills/research/SKILL.md b/plugins/core-copilot/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-copilot/skills/research/SKILL.md +++ b/plugins/core-copilot/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-cursor-light/skills/research/README.md b/plugins/core-cursor-light/skills/research/README.md index fb36adf58..e7efb0b31 100644 --- a/plugins/core-cursor-light/skills/research/README.md +++ b/plugins/core-cursor-light/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-cursor-light/skills/research/SKILL.md b/plugins/core-cursor-light/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-cursor-light/skills/research/SKILL.md +++ b/plugins/core-cursor-light/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md b/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md index fb36adf58..e7efb0b31 100644 --- a/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md +++ b/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md b/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md +++ b/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-cursor-standalone/.cursor/skills/research/README.md b/plugins/core-cursor-standalone/.cursor/skills/research/README.md index fb36adf58..e7efb0b31 100644 --- a/plugins/core-cursor-standalone/.cursor/skills/research/README.md +++ b/plugins/core-cursor-standalone/.cursor/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md b/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md +++ b/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: diff --git a/plugins/core-cursor/skills/research/README.md b/plugins/core-cursor/skills/research/README.md index fb36adf58..e7efb0b31 100644 --- a/plugins/core-cursor/skills/research/README.md +++ b/plugins/core-cursor/skills/research/README.md @@ -10,7 +10,7 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel-slice proposals returned to the orchestrator) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules - "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. diff --git a/plugins/core-cursor/skills/research/SKILL.md b/plugins/core-cursor/skills/research/SKILL.md index 2941a0798..eedd389ce 100644 --- a/plugins/core-cursor/skills/research/SKILL.md +++ b/plugins/core-cursor/skills/research/SKILL.md @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it directly in the current researcher context (subagents cannot dispatch subagents; the workflow owns dispatch) +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). - Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -45,7 +45,7 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the top-level orchestrator; do NOT spawn subagents from here (subagents cannot dispatch subagents). +- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. - Use synthesis and comparison approach Enforcement rules for the generated research prompt: From 0b3818eb7101d001560b8e418a3811605e0aea45 Mon Sep 17 00:00:00 2001 From: ab Date: Thu, 15 Oct 2026 18:50:01 +0300 Subject: [PATCH 3/5] fix(instructions): resolve review findings on reduced-route and research contracts MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - #356 follow-through: the reduced route now defines its substitutes IN the phase files the agent loads — modernization-flow-analysis.md and modernization-flow-mapping.md carry explicit reduced-route exceptions on their prerequisites and MUST re-read lines; the sequential-execution note is qualified; the state-file template gains a Reduced Route field so the exception is auditable; Phase 7 is explicitly skipped on the reduced route (Phase 6 single-project note is its substitute); 'final-review' wording corrected to 'single-project mapping scope'. - research SKILL.md: removed the capability-aware source line that duplicated enforcement rule 2; role wording aligned ('execute that prompt — never ad-hoc research'). - research README: self-contradictory meta-prompting line rewritten to the caller-scoped contract; stale line citation corrected (research-flow.md:50). Regenerated across all 14 plugin targets; 797 generator tests pass. --- instructions/r3/core/skills/research/README.md | 6 +++--- instructions/r3/core/skills/research/SKILL.md | 3 +-- .../r3/core/workflows/modernization-flow-analysis.md | 4 ++-- .../r3/core/workflows/modernization-flow-mapping.md | 6 +++--- instructions/r3/core/workflows/modernization-flow.md | 8 ++++++-- .../skills/modernization-flow/SKILL.md | 8 ++++++-- .../phases/modernization-flow-analysis.md | 4 ++-- .../phases/modernization-flow-mapping.md | 6 +++--- plugins/core-antigravity-light/skills/research/README.md | 6 +++--- plugins/core-antigravity-light/skills/research/SKILL.md | 3 +-- .../core-antigravity/skills/modernization-flow/SKILL.md | 8 ++++++-- .../phases/modernization-flow-analysis.md | 4 ++-- .../phases/modernization-flow-mapping.md | 6 +++--- plugins/core-antigravity/skills/research/README.md | 6 +++--- plugins/core-antigravity/skills/research/SKILL.md | 3 +-- plugins/core-claude-light/skills/research/README.md | 6 +++--- plugins/core-claude-light/skills/research/SKILL.md | 3 +-- .../workflows/modernization-flow-analysis.md | 4 ++-- .../workflows/modernization-flow-mapping.md | 6 +++--- plugins/core-claude-light/workflows/modernization-flow.md | 8 ++++++-- plugins/core-claude/skills/research/README.md | 6 +++--- plugins/core-claude/skills/research/SKILL.md | 3 +-- .../core-claude/workflows/modernization-flow-analysis.md | 4 ++-- .../core-claude/workflows/modernization-flow-mapping.md | 6 +++--- plugins/core-claude/workflows/modernization-flow.md | 8 ++++++-- .../.agents/skills/modernization-flow/SKILL.md | 8 ++++++-- .../phases/modernization-flow-analysis.md | 4 ++-- .../phases/modernization-flow-mapping.md | 6 +++--- .../core-codex-light/.agents/skills/research/README.md | 6 +++--- plugins/core-codex-light/.agents/skills/research/SKILL.md | 3 +-- .../core-codex/.agents/skills/modernization-flow/SKILL.md | 8 ++++++-- .../phases/modernization-flow-analysis.md | 4 ++-- .../phases/modernization-flow-mapping.md | 6 +++--- plugins/core-codex/.agents/skills/research/README.md | 6 +++--- plugins/core-codex/.agents/skills/research/SKILL.md | 3 +-- .../commands/modernization-flow-analysis.md | 4 ++-- .../commands/modernization-flow-mapping.md | 6 +++--- plugins/core-copilot-light/commands/modernization-flow.md | 8 ++++++-- plugins/core-copilot-light/skills/research/README.md | 6 +++--- plugins/core-copilot-light/skills/research/SKILL.md | 3 +-- .../.github/prompts/modernization-flow-analysis.prompt.md | 4 ++-- .../.github/prompts/modernization-flow-mapping.prompt.md | 6 +++--- .../.github/prompts/modernization-flow.prompt.md | 8 ++++++-- .../.github/skills/research/README.md | 6 +++--- .../.github/skills/research/SKILL.md | 3 +-- .../.github/prompts/modernization-flow-analysis.prompt.md | 4 ++-- .../.github/prompts/modernization-flow-mapping.prompt.md | 6 +++--- .../.github/prompts/modernization-flow.prompt.md | 8 ++++++-- .../.github/skills/research/README.md | 6 +++--- .../.github/skills/research/SKILL.md | 3 +-- .../core-copilot/commands/modernization-flow-analysis.md | 4 ++-- .../core-copilot/commands/modernization-flow-mapping.md | 6 +++--- plugins/core-copilot/commands/modernization-flow.md | 8 ++++++-- plugins/core-copilot/skills/research/README.md | 6 +++--- plugins/core-copilot/skills/research/SKILL.md | 3 +-- .../commands/modernization-flow-analysis.md | 4 ++-- .../commands/modernization-flow-mapping.md | 6 +++--- plugins/core-cursor-light/commands/modernization-flow.md | 8 ++++++-- plugins/core-cursor-light/skills/research/README.md | 6 +++--- plugins/core-cursor-light/skills/research/SKILL.md | 3 +-- .../.cursor/commands/modernization-flow-analysis.md | 4 ++-- .../.cursor/commands/modernization-flow-mapping.md | 6 +++--- .../.cursor/commands/modernization-flow.md | 8 ++++++-- .../.cursor/skills/research/README.md | 6 +++--- .../.cursor/skills/research/SKILL.md | 3 +-- .../.cursor/commands/modernization-flow-analysis.md | 4 ++-- .../.cursor/commands/modernization-flow-mapping.md | 6 +++--- .../.cursor/commands/modernization-flow.md | 8 ++++++-- .../.cursor/skills/research/README.md | 6 +++--- .../.cursor/skills/research/SKILL.md | 3 +-- .../core-cursor/commands/modernization-flow-analysis.md | 4 ++-- .../core-cursor/commands/modernization-flow-mapping.md | 6 +++--- plugins/core-cursor/commands/modernization-flow.md | 8 ++++++-- plugins/core-cursor/skills/research/README.md | 6 +++--- plugins/core-cursor/skills/research/SKILL.md | 3 +-- 75 files changed, 225 insertions(+), 180 deletions(-) diff --git a/instructions/r3/core/skills/research/README.md b/instructions/r3/core/skills/research/README.md index 9a3fe7493..65221f7da 100644 --- a/instructions/r3/core/skills/research/README.md +++ b/instructions/r3/core/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`workflows/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `workflows/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `workflows/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `workflows/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `workflows/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `workflows/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `workflows/self-help-flow.md` (as a cautionary example). diff --git a/instructions/r3/core/skills/research/SKILL.md b/instructions/r3/core/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/instructions/r3/core/skills/research/SKILL.md +++ b/instructions/r3/core/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/instructions/r3/core/workflows/modernization-flow-analysis.md b/instructions/r3/core/workflows/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/instructions/r3/core/workflows/modernization-flow-analysis.md +++ b/instructions/r3/core/workflows/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/instructions/r3/core/workflows/modernization-flow-mapping.md b/instructions/r3/core/workflows/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/instructions/r3/core/workflows/modernization-flow-mapping.md +++ b/instructions/r3/core/workflows/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/instructions/r3/core/workflows/modernization-flow.md b/instructions/r3/core/workflows/modernization-flow.md index 621041514..0e818caba 100644 --- a/instructions/r3/core/workflows/modernization-flow.md +++ b/instructions/r3/core/workflows/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md b/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md index 0c43753b6..295543fdc 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md @@ -28,7 +28,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -112,6 +112,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -125,7 +129,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-analysis.md b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-analysis.md index 2b64084b0..2b0a8ccb9 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-analysis.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-analysis.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -17,7 +17,7 @@ ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md index b3271bd9d..1b7717d54 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-antigravity-light/skills/research/README.md b/plugins/core-antigravity-light/skills/research/README.md index 79b9cced5..b036c061f 100644 --- a/plugins/core-antigravity-light/skills/research/README.md +++ b/plugins/core-antigravity-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`skills/research-flow/SKILL.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `skills/research-flow/SKILL.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `skills/self-help-flow/SKILL.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `skills/research-flow/SKILL.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `skills/self-help-flow/SKILL.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `skills/research-flow/SKILL.md`, `skills/orchestration/assets/o-team-manager.md`, and `skills/self-help-flow/SKILL.md` (as a cautionary example). diff --git a/plugins/core-antigravity-light/skills/research/SKILL.md b/plugins/core-antigravity-light/skills/research/SKILL.md index a90cc66b1..b77aa7d51 100644 --- a/plugins/core-antigravity-light/skills/research/SKILL.md +++ b/plugins/core-antigravity-light/skills/research/SKILL.md @@ -7,7 +7,7 @@ description: "To run systematic deep research via meta-prompting — grounded re -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -19,7 +19,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-antigravity/skills/modernization-flow/SKILL.md b/plugins/core-antigravity/skills/modernization-flow/SKILL.md index 0c43753b6..295543fdc 100644 --- a/plugins/core-antigravity/skills/modernization-flow/SKILL.md +++ b/plugins/core-antigravity/skills/modernization-flow/SKILL.md @@ -28,7 +28,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -112,6 +112,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -125,7 +129,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-analysis.md b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-analysis.md index 2b64084b0..2b0a8ccb9 100644 --- a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-analysis.md +++ b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-analysis.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -17,7 +17,7 @@ ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md index b3271bd9d..1b7717d54 100644 --- a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-antigravity/skills/research/README.md b/plugins/core-antigravity/skills/research/README.md index 79b9cced5..b036c061f 100644 --- a/plugins/core-antigravity/skills/research/README.md +++ b/plugins/core-antigravity/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`skills/research-flow/SKILL.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `skills/research-flow/SKILL.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `skills/self-help-flow/SKILL.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `skills/research-flow/SKILL.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `skills/self-help-flow/SKILL.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `skills/research-flow/SKILL.md`, `skills/orchestration/assets/o-team-manager.md`, and `skills/self-help-flow/SKILL.md` (as a cautionary example). diff --git a/plugins/core-antigravity/skills/research/SKILL.md b/plugins/core-antigravity/skills/research/SKILL.md index a90cc66b1..b77aa7d51 100644 --- a/plugins/core-antigravity/skills/research/SKILL.md +++ b/plugins/core-antigravity/skills/research/SKILL.md @@ -7,7 +7,7 @@ description: "To run systematic deep research via meta-prompting — grounded re -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -19,7 +19,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-claude-light/skills/research/README.md b/plugins/core-claude-light/skills/research/README.md index 9a3fe7493..65221f7da 100644 --- a/plugins/core-claude-light/skills/research/README.md +++ b/plugins/core-claude-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`workflows/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `workflows/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `workflows/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `workflows/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `workflows/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `workflows/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `workflows/self-help-flow.md` (as a cautionary example). diff --git a/plugins/core-claude-light/skills/research/SKILL.md b/plugins/core-claude-light/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-claude-light/skills/research/SKILL.md +++ b/plugins/core-claude-light/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-claude-light/workflows/modernization-flow-analysis.md b/plugins/core-claude-light/workflows/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-claude-light/workflows/modernization-flow-analysis.md +++ b/plugins/core-claude-light/workflows/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-claude-light/workflows/modernization-flow-mapping.md b/plugins/core-claude-light/workflows/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-claude-light/workflows/modernization-flow-mapping.md +++ b/plugins/core-claude-light/workflows/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-claude-light/workflows/modernization-flow.md b/plugins/core-claude-light/workflows/modernization-flow.md index 621041514..0e818caba 100644 --- a/plugins/core-claude-light/workflows/modernization-flow.md +++ b/plugins/core-claude-light/workflows/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-claude/skills/research/README.md b/plugins/core-claude/skills/research/README.md index 9a3fe7493..65221f7da 100644 --- a/plugins/core-claude/skills/research/README.md +++ b/plugins/core-claude/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`workflows/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `workflows/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `workflows/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `workflows/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `workflows/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `workflows/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `workflows/self-help-flow.md` (as a cautionary example). diff --git a/plugins/core-claude/skills/research/SKILL.md b/plugins/core-claude/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-claude/skills/research/SKILL.md +++ b/plugins/core-claude/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-claude/workflows/modernization-flow-analysis.md b/plugins/core-claude/workflows/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-claude/workflows/modernization-flow-analysis.md +++ b/plugins/core-claude/workflows/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-claude/workflows/modernization-flow-mapping.md b/plugins/core-claude/workflows/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-claude/workflows/modernization-flow-mapping.md +++ b/plugins/core-claude/workflows/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-claude/workflows/modernization-flow.md b/plugins/core-claude/workflows/modernization-flow.md index 621041514..0e818caba 100644 --- a/plugins/core-claude/workflows/modernization-flow.md +++ b/plugins/core-claude/workflows/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md b/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md index d6c1bd663..e70955c87 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md index 2b64084b0..2b0a8ccb9 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -17,7 +17,7 @@ ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md index b3271bd9d..1b7717d54 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-codex-light/.agents/skills/research/README.md b/plugins/core-codex-light/.agents/skills/research/README.md index 79b9cced5..b036c061f 100644 --- a/plugins/core-codex-light/.agents/skills/research/README.md +++ b/plugins/core-codex-light/.agents/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`skills/research-flow/SKILL.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `skills/research-flow/SKILL.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `skills/self-help-flow/SKILL.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `skills/research-flow/SKILL.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `skills/self-help-flow/SKILL.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `skills/research-flow/SKILL.md`, `skills/orchestration/assets/o-team-manager.md`, and `skills/self-help-flow/SKILL.md` (as a cautionary example). diff --git a/plugins/core-codex-light/.agents/skills/research/SKILL.md b/plugins/core-codex-light/.agents/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-codex-light/.agents/skills/research/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md b/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md index d6c1bd663..e70955c87 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md index 2b64084b0..2b0a8ccb9 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -17,7 +17,7 @@ ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md index b3271bd9d..1b7717d54 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-codex/.agents/skills/research/README.md b/plugins/core-codex/.agents/skills/research/README.md index 79b9cced5..b036c061f 100644 --- a/plugins/core-codex/.agents/skills/research/README.md +++ b/plugins/core-codex/.agents/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`skills/research-flow/SKILL.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `skills/research-flow/SKILL.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `skills/self-help-flow/SKILL.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `skills/research-flow/SKILL.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `skills/self-help-flow/SKILL.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `skills/research-flow/SKILL.md`, `skills/orchestration/assets/o-team-manager.md`, and `skills/self-help-flow/SKILL.md` (as a cautionary example). diff --git a/plugins/core-codex/.agents/skills/research/SKILL.md b/plugins/core-codex/.agents/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-codex/.agents/skills/research/SKILL.md +++ b/plugins/core-codex/.agents/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-copilot-light/commands/modernization-flow-analysis.md b/plugins/core-copilot-light/commands/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-copilot-light/commands/modernization-flow-analysis.md +++ b/plugins/core-copilot-light/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-copilot-light/commands/modernization-flow-mapping.md b/plugins/core-copilot-light/commands/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-copilot-light/commands/modernization-flow-mapping.md +++ b/plugins/core-copilot-light/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-light/commands/modernization-flow.md b/plugins/core-copilot-light/commands/modernization-flow.md index 621041514..0e818caba 100644 --- a/plugins/core-copilot-light/commands/modernization-flow.md +++ b/plugins/core-copilot-light/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-copilot-light/skills/research/README.md b/plugins/core-copilot-light/skills/research/README.md index 4b2131ae3..f4c58de1e 100644 --- a/plugins/core-copilot-light/skills/research/README.md +++ b/plugins/core-copilot-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.agent.md`) is the skill-level tracker; `research-flow-state.md` (`commands/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.agent.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.agent.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.agent.md`, `commands/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `commands/self-help-flow.md` (as a cautionary example). diff --git a/plugins/core-copilot-light/skills/research/SKILL.md b/plugins/core-copilot-light/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-copilot-light/skills/research/SKILL.md +++ b/plugins/core-copilot-light/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-analysis.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-analysis.prompt.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-analysis.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-analysis.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md index 621041514..0e818caba 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-copilot-standalone-light/.github/skills/research/README.md b/plugins/core-copilot-standalone-light/.github/skills/research/README.md index f7b492df5..7af890869 100644 --- a/plugins/core-copilot-standalone-light/.github/skills/research/README.md +++ b/plugins/core-copilot-standalone-light/.github/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.agent.md`) is the skill-level tracker; `research-flow-state.md` (`prompts/research-flow.prompt.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.agent.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `prompts/research-flow.prompt.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `prompts/self-help-flow.prompt.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.agent.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `prompts/research-flow.prompt.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `prompts/self-help-flow.prompt.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.agent.md`, `prompts/research-flow.prompt.md`, `skills/orchestration/assets/o-team-manager.md`, and `prompts/self-help-flow.prompt.md` (as a cautionary example). diff --git a/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md b/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md +++ b/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-analysis.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-analysis.prompt.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-analysis.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-analysis.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md index 621041514..0e818caba 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-copilot-standalone/.github/skills/research/README.md b/plugins/core-copilot-standalone/.github/skills/research/README.md index f7b492df5..7af890869 100644 --- a/plugins/core-copilot-standalone/.github/skills/research/README.md +++ b/plugins/core-copilot-standalone/.github/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.agent.md`) is the skill-level tracker; `research-flow-state.md` (`prompts/research-flow.prompt.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.agent.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `prompts/research-flow.prompt.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `prompts/self-help-flow.prompt.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.agent.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `prompts/research-flow.prompt.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `prompts/self-help-flow.prompt.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.agent.md`, `prompts/research-flow.prompt.md`, `skills/orchestration/assets/o-team-manager.md`, and `prompts/self-help-flow.prompt.md` (as a cautionary example). diff --git a/plugins/core-copilot-standalone/.github/skills/research/SKILL.md b/plugins/core-copilot-standalone/.github/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-copilot-standalone/.github/skills/research/SKILL.md +++ b/plugins/core-copilot-standalone/.github/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-copilot/commands/modernization-flow-analysis.md b/plugins/core-copilot/commands/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-copilot/commands/modernization-flow-analysis.md +++ b/plugins/core-copilot/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-copilot/commands/modernization-flow-mapping.md b/plugins/core-copilot/commands/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-copilot/commands/modernization-flow-mapping.md +++ b/plugins/core-copilot/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot/commands/modernization-flow.md b/plugins/core-copilot/commands/modernization-flow.md index 621041514..0e818caba 100644 --- a/plugins/core-copilot/commands/modernization-flow.md +++ b/plugins/core-copilot/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-copilot/skills/research/README.md b/plugins/core-copilot/skills/research/README.md index 4b2131ae3..f4c58de1e 100644 --- a/plugins/core-copilot/skills/research/README.md +++ b/plugins/core-copilot/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.agent.md`) is the skill-level tracker; `research-flow-state.md` (`commands/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.agent.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.agent.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.agent.md`, `commands/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `commands/self-help-flow.md` (as a cautionary example). diff --git a/plugins/core-copilot/skills/research/SKILL.md b/plugins/core-copilot/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-copilot/skills/research/SKILL.md +++ b/plugins/core-copilot/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-cursor-light/commands/modernization-flow-analysis.md b/plugins/core-cursor-light/commands/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-cursor-light/commands/modernization-flow-analysis.md +++ b/plugins/core-cursor-light/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-cursor-light/commands/modernization-flow-mapping.md b/plugins/core-cursor-light/commands/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-cursor-light/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-light/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-light/commands/modernization-flow.md b/plugins/core-cursor-light/commands/modernization-flow.md index 621041514..0e818caba 100644 --- a/plugins/core-cursor-light/commands/modernization-flow.md +++ b/plugins/core-cursor-light/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-cursor-light/skills/research/README.md b/plugins/core-cursor-light/skills/research/README.md index e7efb0b31..942562b7f 100644 --- a/plugins/core-cursor-light/skills/research/README.md +++ b/plugins/core-cursor-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`commands/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `commands/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `commands/self-help-flow.md` (as a cautionary example). diff --git a/plugins/core-cursor-light/skills/research/SKILL.md b/plugins/core-cursor-light/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-cursor-light/skills/research/SKILL.md +++ b/plugins/core-cursor-light/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-analysis.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-analysis.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md index 621041514..0e818caba 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md b/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md index e7efb0b31..942562b7f 100644 --- a/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md +++ b/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`commands/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `commands/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `commands/self-help-flow.md` (as a cautionary example). diff --git a/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md b/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md +++ b/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-analysis.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-analysis.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md index 621041514..0e818caba 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-cursor-standalone/.cursor/skills/research/README.md b/plugins/core-cursor-standalone/.cursor/skills/research/README.md index e7efb0b31..942562b7f 100644 --- a/plugins/core-cursor-standalone/.cursor/skills/research/README.md +++ b/plugins/core-cursor-standalone/.cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`commands/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `commands/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `commands/self-help-flow.md` (as a cautionary example). diff --git a/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md b/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md +++ b/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly diff --git a/plugins/core-cursor/commands/modernization-flow-analysis.md b/plugins/core-cursor/commands/modernization-flow-analysis.md index 3e1710fd4..65ed95af8 100644 --- a/plugins/core-cursor/commands/modernization-flow-analysis.md +++ b/plugins/core-cursor/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first +1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created) +**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor/commands/modernization-flow-mapping.md index e1e92cc65..1094fd02b 100644 --- a/plugins/core-cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first +1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis) +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor/commands/modernization-flow.md b/plugins/core-cursor/commands/modernization-flow.md index 621041514..0e818caba 100644 --- a/plugins/core-cursor/commands/modernization-flow.md +++ b/plugins/core-cursor/commands/modernization-flow.md @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project final-review contract — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,6 +114,10 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started +## Reduced Route + +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). + ## Phase Details ### Phase 1 @@ -127,7 +131,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order +- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-cursor/skills/research/README.md b/plugins/core-cursor/skills/research/README.md index e7efb0b31..942562b7f 100644 --- a/plugins/core-cursor/skills/research/README.md +++ b/plugins/core-cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute it directly in the current researcher context — never research directly, and never dispatch nested subagents (the workflow owns dispatch). +Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -13,7 +13,7 @@ Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. +- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. @@ -25,7 +25,7 @@ Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior rese - `description` is one dense sentence under the ~25-token budget ("To run systematic deep research via meta-prompting — grounded references, incremental tracking, self-validation."). - XML section names ``, ``, ``, ``, `` are structural anchors that other tooling may parse. - Two distinct, similarly-named state files: `research-state.md` (this skill, `agents/researcher.md`) is the skill-level tracker; `research-flow-state.md` (`commands/research-flow.md`) is the workflow-level tracker — do not rename either without checking both call sites, and do not assume they are the same file. -- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:51` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. +- Inbound couplings: `agents/researcher.md:33,40` ("Apply `research` skill." / "USE SKILL research") — the subagent contractually required to run this skill; `commands/research-flow.md:50` ("Required skills: `research`", phase 3 `execute_research`) — the sanctioned entry point; `skills/orchestration/assets/o-team-manager.md:9` ("USE SKILL `research` for external knowledge if needed") — an ad hoc invocation path outside `research-flow`; `commands/self-help-flow.md:53` — documents direct `/research` invocation as bypassing the structured workflow. ## Editing guide Safe to edit: wording inside the "Research rules" / "Enforcement rules for the generated research prompt" bullet lists, as long as the meta-prompting split survives (this skill governs how research is carried out; the crafted prompt separately carries its own 5 enforcement rules). Handle with care: frontmatter (`name`, `description`), the two state-file names, and the `docs/-research.md` / `research-prompt.md` output paths, since `research-flow.md` and `researcher.md` branch on these exact strings. Referenced by: `agents/researcher.md`, `commands/research-flow.md`, `skills/orchestration/assets/o-team-manager.md`, and `commands/self-help-flow.md` (as a cautionary example). diff --git a/plugins/core-cursor/skills/research/SKILL.md b/plugins/core-cursor/skills/research/SKILL.md index eedd389ce..eefd47ae6 100644 --- a/plugins/core-cursor/skills/research/SKILL.md +++ b/plugins/core-cursor/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. @@ -22,7 +22,6 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed - Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). -- Capability-aware source tooling: prefer DeepWiki and Context7 when those integrations are available in the current environment; otherwise fall back to project documentation, vendor docs, and web search, and name the source actually used. - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly From c6ca184019ae171ce02d08ccd853c3c12d353cda Mon Sep 17 00:00:00 2001 From: ab Date: Thu, 22 Oct 2026 22:42:49 +0300 Subject: [PATCH 4/5] fix(instructions): close remaining reduced-route gaps in modernization flow - Phase 8 (implement) prerequisite now carries the reduced-route exception: Phase 7 is skipped there, so Phase 6 + the single-project scope note stand in for the cross-project review. - "DO NOT SKIP PHASES" / "Do not skip phases!" in Critical Requirements now name the reduced route as the only exception. - The single-project scope note has a defined home: the orchestrator writes it in the state file's Reduced Route section before Phase 6; Phase 6 prerequisites and its re-read requirement point there. Regenerated across all 14 plugin targets; 797 generator tests pass. Co-Authored-By: Claude Opus 5.5 --- .../r3/core/workflows/modernization-flow-implement.md | 2 +- .../r3/core/workflows/modernization-flow-mapping.md | 4 ++-- instructions/r3/core/workflows/modernization-flow.md | 8 ++++---- .../skills/modernization-flow/SKILL.md | 8 ++++---- .../phases/modernization-flow-implement.md | 2 +- .../phases/modernization-flow-mapping.md | 4 ++-- .../core-antigravity/skills/modernization-flow/SKILL.md | 8 ++++---- .../phases/modernization-flow-implement.md | 2 +- .../phases/modernization-flow-mapping.md | 4 ++-- .../workflows/modernization-flow-implement.md | 2 +- .../workflows/modernization-flow-mapping.md | 4 ++-- plugins/core-claude-light/workflows/modernization-flow.md | 8 ++++---- .../core-claude/workflows/modernization-flow-implement.md | 2 +- .../core-claude/workflows/modernization-flow-mapping.md | 4 ++-- plugins/core-claude/workflows/modernization-flow.md | 8 ++++---- .../.agents/skills/modernization-flow/SKILL.md | 8 ++++---- .../phases/modernization-flow-implement.md | 2 +- .../phases/modernization-flow-mapping.md | 4 ++-- .../core-codex/.agents/skills/modernization-flow/SKILL.md | 8 ++++---- .../phases/modernization-flow-implement.md | 2 +- .../phases/modernization-flow-mapping.md | 4 ++-- .../commands/modernization-flow-implement.md | 2 +- .../commands/modernization-flow-mapping.md | 4 ++-- plugins/core-copilot-light/commands/modernization-flow.md | 8 ++++---- .../prompts/modernization-flow-implement.prompt.md | 2 +- .../.github/prompts/modernization-flow-mapping.prompt.md | 4 ++-- .../.github/prompts/modernization-flow.prompt.md | 8 ++++---- .../prompts/modernization-flow-implement.prompt.md | 2 +- .../.github/prompts/modernization-flow-mapping.prompt.md | 4 ++-- .../.github/prompts/modernization-flow.prompt.md | 8 ++++---- .../core-copilot/commands/modernization-flow-implement.md | 2 +- .../core-copilot/commands/modernization-flow-mapping.md | 4 ++-- plugins/core-copilot/commands/modernization-flow.md | 8 ++++---- .../commands/modernization-flow-implement.md | 2 +- .../commands/modernization-flow-mapping.md | 4 ++-- plugins/core-cursor-light/commands/modernization-flow.md | 8 ++++---- .../.cursor/commands/modernization-flow-implement.md | 2 +- .../.cursor/commands/modernization-flow-mapping.md | 4 ++-- .../.cursor/commands/modernization-flow.md | 8 ++++---- .../.cursor/commands/modernization-flow-implement.md | 2 +- .../.cursor/commands/modernization-flow-mapping.md | 4 ++-- .../.cursor/commands/modernization-flow.md | 8 ++++---- .../core-cursor/commands/modernization-flow-implement.md | 2 +- .../core-cursor/commands/modernization-flow-mapping.md | 4 ++-- plugins/core-cursor/commands/modernization-flow.md | 8 ++++---- 45 files changed, 105 insertions(+), 105 deletions(-) diff --git a/instructions/r3/core/workflows/modernization-flow-implement.md b/instructions/r3/core/workflows/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/instructions/r3/core/workflows/modernization-flow-implement.md +++ b/instructions/r3/core/workflows/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/instructions/r3/core/workflows/modernization-flow-mapping.md b/instructions/r3/core/workflows/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/instructions/r3/core/workflows/modernization-flow-mapping.md +++ b/instructions/r3/core/workflows/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/instructions/r3/core/workflows/modernization-flow.md b/instructions/r3/core/workflows/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/instructions/r3/core/workflows/modernization-flow.md +++ b/instructions/r3/core/workflows/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md b/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md index 295543fdc..cfe052df8 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md @@ -12,10 +12,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -28,7 +28,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,7 +114,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-implement.md b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-implement.md index 5a411eaab..55a50e2e7 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-implement.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-implement.md @@ -3,7 +3,7 @@ ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md index 1b7717d54..ff77204ed 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-antigravity/skills/modernization-flow/SKILL.md b/plugins/core-antigravity/skills/modernization-flow/SKILL.md index 295543fdc..cfe052df8 100644 --- a/plugins/core-antigravity/skills/modernization-flow/SKILL.md +++ b/plugins/core-antigravity/skills/modernization-flow/SKILL.md @@ -12,10 +12,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -28,7 +28,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,7 +114,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-implement.md b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-implement.md index 5a411eaab..55a50e2e7 100644 --- a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-implement.md +++ b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-implement.md @@ -3,7 +3,7 @@ ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md index 1b7717d54..ff77204ed 100644 --- a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-claude-light/workflows/modernization-flow-implement.md b/plugins/core-claude-light/workflows/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-claude-light/workflows/modernization-flow-implement.md +++ b/plugins/core-claude-light/workflows/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-claude-light/workflows/modernization-flow-mapping.md b/plugins/core-claude-light/workflows/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-claude-light/workflows/modernization-flow-mapping.md +++ b/plugins/core-claude-light/workflows/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-claude-light/workflows/modernization-flow.md b/plugins/core-claude-light/workflows/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-claude-light/workflows/modernization-flow.md +++ b/plugins/core-claude-light/workflows/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-claude/workflows/modernization-flow-implement.md b/plugins/core-claude/workflows/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-claude/workflows/modernization-flow-implement.md +++ b/plugins/core-claude/workflows/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-claude/workflows/modernization-flow-mapping.md b/plugins/core-claude/workflows/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-claude/workflows/modernization-flow-mapping.md +++ b/plugins/core-claude/workflows/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-claude/workflows/modernization-flow.md b/plugins/core-claude/workflows/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-claude/workflows/modernization-flow.md +++ b/plugins/core-claude/workflows/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md b/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md index e70955c87..fa1dbfdd8 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-implement.md b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-implement.md index 5a411eaab..55a50e2e7 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-implement.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-implement.md @@ -3,7 +3,7 @@ ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md index 1b7717d54..ff77204ed 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md b/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md index e70955c87..fa1dbfdd8 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-implement.md b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-implement.md index 5a411eaab..55a50e2e7 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-implement.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-implement.md @@ -3,7 +3,7 @@ ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md index 1b7717d54..ff77204ed 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-light/commands/modernization-flow-implement.md b/plugins/core-copilot-light/commands/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-copilot-light/commands/modernization-flow-implement.md +++ b/plugins/core-copilot-light/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-copilot-light/commands/modernization-flow-mapping.md b/plugins/core-copilot-light/commands/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-copilot-light/commands/modernization-flow-mapping.md +++ b/plugins/core-copilot-light/commands/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-light/commands/modernization-flow.md b/plugins/core-copilot-light/commands/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-copilot-light/commands/modernization-flow.md +++ b/plugins/core-copilot-light/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-implement.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-implement.prompt.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-implement.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-implement.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-implement.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-implement.prompt.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-implement.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-implement.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-copilot/commands/modernization-flow-implement.md b/plugins/core-copilot/commands/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-copilot/commands/modernization-flow-implement.md +++ b/plugins/core-copilot/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-copilot/commands/modernization-flow-mapping.md b/plugins/core-copilot/commands/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-copilot/commands/modernization-flow-mapping.md +++ b/plugins/core-copilot/commands/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot/commands/modernization-flow.md b/plugins/core-copilot/commands/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-copilot/commands/modernization-flow.md +++ b/plugins/core-copilot/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-cursor-light/commands/modernization-flow-implement.md b/plugins/core-cursor-light/commands/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-cursor-light/commands/modernization-flow-implement.md +++ b/plugins/core-cursor-light/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-cursor-light/commands/modernization-flow-mapping.md b/plugins/core-cursor-light/commands/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-cursor-light/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-light/commands/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-light/commands/modernization-flow.md b/plugins/core-cursor-light/commands/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-cursor-light/commands/modernization-flow.md +++ b/plugins/core-cursor-light/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-implement.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-implement.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-implement.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-implement.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details diff --git a/plugins/core-cursor/commands/modernization-flow-implement.md b/plugins/core-cursor/commands/modernization-flow-implement.md index e6cb3281e..57460b63d 100644 --- a/plugins/core-cursor/commands/modernization-flow-implement.md +++ b/plugins/core-cursor/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review) +- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor/commands/modernization-flow-mapping.md index 1094fd02b..38b84b8e4 100644 --- a/plugins/core-cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor/commands/modernization-flow-mapping.md @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; substitute an explicit note that mapping is scoped to this project alone. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note instead, per the PREREQUISITES exception) +- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor/commands/modernization-flow.md b/plugins/core-cursor/commands/modernization-flow.md index 0e818caba..814f31ec0 100644 --- a/plugins/core-cursor/commands/modernization-flow.md +++ b/plugins/core-cursor/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases! +- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit note that mapping is scoped to this project alone. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -116,7 +116,7 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin ## Reduced Route -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Phase 7 skipped (substituted by the Phase 6 single-project note). +Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). ## Phase Details From d1e237ddedfe7199ab7eb5580dbd68db7c37d70d Mon Sep 17 00:00:00 2001 From: isolomatov-gd Date: Thu, 29 Oct 2026 15:51:05 -0400 Subject: [PATCH 5/5] fix(instructions): keep #368 contract fixes minimal Revert the research skill dispatch rewrite (#354): subagents can spawn subagents, so the original rule stands. Reduce #355 to "when available". Reduce the modernization reduced-route exceptions (#356) to one clause per prerequisite. Regenerate plugins. Co-Authored-By: Claude Opus 5.5 --- instructions/r3/core/skills/research/README.md | 8 ++++---- instructions/r3/core/skills/research/SKILL.md | 8 ++++---- .../r3/core/workflows/modernization-flow-analysis.md | 4 ++-- .../core/workflows/modernization-flow-implement.md | 2 +- .../r3/core/workflows/modernization-flow-mapping.md | 6 +++--- instructions/r3/core/workflows/modernization-flow.md | 12 ++++-------- instructions/r3/core/workflows/research-flow.md | 2 +- .../skills/modernization-flow/SKILL.md | 12 ++++-------- .../phases/modernization-flow-analysis.md | 4 ++-- .../phases/modernization-flow-implement.md | 2 +- .../phases/modernization-flow-mapping.md | 6 +++--- .../skills/research-flow/SKILL.md | 2 +- .../core-antigravity-light/skills/research/README.md | 8 ++++---- .../core-antigravity-light/skills/research/SKILL.md | 8 ++++---- .../skills/modernization-flow/SKILL.md | 12 ++++-------- .../phases/modernization-flow-analysis.md | 4 ++-- .../phases/modernization-flow-implement.md | 2 +- .../phases/modernization-flow-mapping.md | 6 +++--- .../core-antigravity/skills/research-flow/SKILL.md | 2 +- plugins/core-antigravity/skills/research/README.md | 8 ++++---- plugins/core-antigravity/skills/research/SKILL.md | 8 ++++---- plugins/core-claude-light/skills/research/README.md | 8 ++++---- plugins/core-claude-light/skills/research/SKILL.md | 8 ++++---- .../workflows/modernization-flow-analysis.md | 4 ++-- .../workflows/modernization-flow-implement.md | 2 +- .../workflows/modernization-flow-mapping.md | 6 +++--- .../workflows/modernization-flow.md | 12 ++++-------- plugins/core-claude-light/workflows/research-flow.md | 2 +- plugins/core-claude/skills/research/README.md | 8 ++++---- plugins/core-claude/skills/research/SKILL.md | 8 ++++---- .../workflows/modernization-flow-analysis.md | 4 ++-- .../workflows/modernization-flow-implement.md | 2 +- .../workflows/modernization-flow-mapping.md | 6 +++--- plugins/core-claude/workflows/modernization-flow.md | 12 ++++-------- plugins/core-claude/workflows/research-flow.md | 2 +- .../.agents/skills/modernization-flow/SKILL.md | 12 ++++-------- .../phases/modernization-flow-analysis.md | 4 ++-- .../phases/modernization-flow-implement.md | 2 +- .../phases/modernization-flow-mapping.md | 6 +++--- .../.agents/skills/research-flow/SKILL.md | 2 +- .../.agents/skills/research/README.md | 8 ++++---- .../.agents/skills/research/SKILL.md | 8 ++++---- .../.agents/skills/modernization-flow/SKILL.md | 12 ++++-------- .../phases/modernization-flow-analysis.md | 4 ++-- .../phases/modernization-flow-implement.md | 2 +- .../phases/modernization-flow-mapping.md | 6 +++--- .../core-codex/.agents/skills/research-flow/SKILL.md | 2 +- plugins/core-codex/.agents/skills/research/README.md | 8 ++++---- plugins/core-codex/.agents/skills/research/SKILL.md | 8 ++++---- .../commands/modernization-flow-analysis.md | 4 ++-- .../commands/modernization-flow-implement.md | 2 +- .../commands/modernization-flow-mapping.md | 6 +++--- .../commands/modernization-flow.md | 12 ++++-------- plugins/core-copilot-light/commands/research-flow.md | 2 +- plugins/core-copilot-light/skills/research/README.md | 8 ++++---- plugins/core-copilot-light/skills/research/SKILL.md | 8 ++++---- .../prompts/modernization-flow-analysis.prompt.md | 4 ++-- .../prompts/modernization-flow-implement.prompt.md | 2 +- .../prompts/modernization-flow-mapping.prompt.md | 6 +++--- .../.github/prompts/modernization-flow.prompt.md | 12 ++++-------- .../.github/prompts/research-flow.prompt.md | 2 +- .../.github/skills/research/README.md | 8 ++++---- .../.github/skills/research/SKILL.md | 8 ++++---- .../prompts/modernization-flow-analysis.prompt.md | 4 ++-- .../prompts/modernization-flow-implement.prompt.md | 2 +- .../prompts/modernization-flow-mapping.prompt.md | 6 +++--- .../.github/prompts/modernization-flow.prompt.md | 12 ++++-------- .../.github/prompts/research-flow.prompt.md | 2 +- .../.github/skills/research/README.md | 8 ++++---- .../.github/skills/research/SKILL.md | 8 ++++---- .../commands/modernization-flow-analysis.md | 4 ++-- .../commands/modernization-flow-implement.md | 2 +- .../commands/modernization-flow-mapping.md | 6 +++--- plugins/core-copilot/commands/modernization-flow.md | 12 ++++-------- plugins/core-copilot/commands/research-flow.md | 2 +- plugins/core-copilot/skills/research/README.md | 8 ++++---- plugins/core-copilot/skills/research/SKILL.md | 8 ++++---- .../commands/modernization-flow-analysis.md | 4 ++-- .../commands/modernization-flow-implement.md | 2 +- .../commands/modernization-flow-mapping.md | 6 +++--- .../core-cursor-light/commands/modernization-flow.md | 12 ++++-------- plugins/core-cursor-light/commands/research-flow.md | 2 +- plugins/core-cursor-light/skills/research/README.md | 8 ++++---- plugins/core-cursor-light/skills/research/SKILL.md | 8 ++++---- .../.cursor/commands/modernization-flow-analysis.md | 4 ++-- .../.cursor/commands/modernization-flow-implement.md | 2 +- .../.cursor/commands/modernization-flow-mapping.md | 6 +++--- .../.cursor/commands/modernization-flow.md | 12 ++++-------- .../.cursor/commands/research-flow.md | 2 +- .../.cursor/skills/research/README.md | 8 ++++---- .../.cursor/skills/research/SKILL.md | 8 ++++---- .../.cursor/commands/modernization-flow-analysis.md | 4 ++-- .../.cursor/commands/modernization-flow-implement.md | 2 +- .../.cursor/commands/modernization-flow-mapping.md | 6 +++--- .../.cursor/commands/modernization-flow.md | 12 ++++-------- .../.cursor/commands/research-flow.md | 2 +- .../.cursor/skills/research/README.md | 8 ++++---- .../.cursor/skills/research/SKILL.md | 8 ++++---- .../commands/modernization-flow-analysis.md | 4 ++-- .../commands/modernization-flow-implement.md | 2 +- .../commands/modernization-flow-mapping.md | 6 +++--- plugins/core-cursor/commands/modernization-flow.md | 12 ++++-------- plugins/core-cursor/commands/research-flow.md | 2 +- plugins/core-cursor/skills/research/README.md | 8 ++++---- plugins/core-cursor/skills/research/SKILL.md | 8 ++++---- 105 files changed, 285 insertions(+), 345 deletions(-) diff --git a/instructions/r3/core/skills/research/README.md b/instructions/r3/core/skills/research/README.md index 65221f7da..5e9af8f6b 100644 --- a/instructions/r3/core/skills/research/README.md +++ b/instructions/r3/core/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/instructions/r3/core/skills/research/SKILL.md b/instructions/r3/core/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/instructions/r3/core/skills/research/SKILL.md +++ b/instructions/r3/core/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/instructions/r3/core/workflows/modernization-flow-analysis.md b/instructions/r3/core/workflows/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/instructions/r3/core/workflows/modernization-flow-analysis.md +++ b/instructions/r3/core/workflows/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/instructions/r3/core/workflows/modernization-flow-implement.md b/instructions/r3/core/workflows/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/instructions/r3/core/workflows/modernization-flow-implement.md +++ b/instructions/r3/core/workflows/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/instructions/r3/core/workflows/modernization-flow-mapping.md b/instructions/r3/core/workflows/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/instructions/r3/core/workflows/modernization-flow-mapping.md +++ b/instructions/r3/core/workflows/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/instructions/r3/core/workflows/modernization-flow.md b/instructions/r3/core/workflows/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/instructions/r3/core/workflows/modernization-flow.md +++ b/instructions/r3/core/workflows/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/instructions/r3/core/workflows/research-flow.md b/instructions/r3/core/workflows/research-flow.md index b7a681499..6d1fe8c67 100644 --- a/instructions/r3/core/workflows/research-flow.md +++ b/instructions/r3/core/workflows/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md b/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md index cfe052df8..c20e2c6c4 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/SKILL.md @@ -12,10 +12,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -28,7 +28,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -112,10 +112,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -129,7 +125,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-analysis.md b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-analysis.md index 2b0a8ccb9..6f1dd855b 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-analysis.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-analysis.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -17,7 +17,7 @@ ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-implement.md b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-implement.md index 55a50e2e7..d525ae744 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-implement.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-implement.md @@ -3,7 +3,7 @@ ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md index ff77204ed..aa7f728ad 100644 --- a/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-antigravity-light/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-antigravity-light/skills/research-flow/SKILL.md b/plugins/core-antigravity-light/skills/research-flow/SKILL.md index 2f6ffeee6..aad1fa0fd 100644 --- a/plugins/core-antigravity-light/skills/research-flow/SKILL.md +++ b/plugins/core-antigravity-light/skills/research-flow/SKILL.md @@ -43,7 +43,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-antigravity-light/skills/research/README.md b/plugins/core-antigravity-light/skills/research/README.md index b036c061f..2dac9d281 100644 --- a/plugins/core-antigravity-light/skills/research/README.md +++ b/plugins/core-antigravity-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-antigravity-light/skills/research/SKILL.md b/plugins/core-antigravity-light/skills/research/SKILL.md index b77aa7d51..ad2d88979 100644 --- a/plugins/core-antigravity-light/skills/research/SKILL.md +++ b/plugins/core-antigravity-light/skills/research/SKILL.md @@ -7,7 +7,7 @@ description: "To run systematic deep research via meta-prompting — grounded re -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -18,7 +18,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -41,13 +41,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-antigravity/skills/modernization-flow/SKILL.md b/plugins/core-antigravity/skills/modernization-flow/SKILL.md index cfe052df8..c20e2c6c4 100644 --- a/plugins/core-antigravity/skills/modernization-flow/SKILL.md +++ b/plugins/core-antigravity/skills/modernization-flow/SKILL.md @@ -12,10 +12,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -28,7 +28,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -112,10 +112,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -129,7 +125,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-analysis.md b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-analysis.md index 2b0a8ccb9..6f1dd855b 100644 --- a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-analysis.md +++ b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-analysis.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -17,7 +17,7 @@ ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-implement.md b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-implement.md index 55a50e2e7..d525ae744 100644 --- a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-implement.md +++ b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-implement.md @@ -3,7 +3,7 @@ ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md index ff77204ed..aa7f728ad 100644 --- a/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-antigravity/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-antigravity/skills/research-flow/SKILL.md b/plugins/core-antigravity/skills/research-flow/SKILL.md index 2f6ffeee6..aad1fa0fd 100644 --- a/plugins/core-antigravity/skills/research-flow/SKILL.md +++ b/plugins/core-antigravity/skills/research-flow/SKILL.md @@ -43,7 +43,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-antigravity/skills/research/README.md b/plugins/core-antigravity/skills/research/README.md index b036c061f..2dac9d281 100644 --- a/plugins/core-antigravity/skills/research/README.md +++ b/plugins/core-antigravity/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-antigravity/skills/research/SKILL.md b/plugins/core-antigravity/skills/research/SKILL.md index b77aa7d51..ad2d88979 100644 --- a/plugins/core-antigravity/skills/research/SKILL.md +++ b/plugins/core-antigravity/skills/research/SKILL.md @@ -7,7 +7,7 @@ description: "To run systematic deep research via meta-prompting — grounded re -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -18,7 +18,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -41,13 +41,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-claude-light/skills/research/README.md b/plugins/core-claude-light/skills/research/README.md index 65221f7da..5e9af8f6b 100644 --- a/plugins/core-claude-light/skills/research/README.md +++ b/plugins/core-claude-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-claude-light/skills/research/SKILL.md b/plugins/core-claude-light/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-claude-light/skills/research/SKILL.md +++ b/plugins/core-claude-light/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-claude-light/workflows/modernization-flow-analysis.md b/plugins/core-claude-light/workflows/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-claude-light/workflows/modernization-flow-analysis.md +++ b/plugins/core-claude-light/workflows/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-claude-light/workflows/modernization-flow-implement.md b/plugins/core-claude-light/workflows/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-claude-light/workflows/modernization-flow-implement.md +++ b/plugins/core-claude-light/workflows/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-claude-light/workflows/modernization-flow-mapping.md b/plugins/core-claude-light/workflows/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-claude-light/workflows/modernization-flow-mapping.md +++ b/plugins/core-claude-light/workflows/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-claude-light/workflows/modernization-flow.md b/plugins/core-claude-light/workflows/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-claude-light/workflows/modernization-flow.md +++ b/plugins/core-claude-light/workflows/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-claude-light/workflows/research-flow.md b/plugins/core-claude-light/workflows/research-flow.md index 2bc2b4816..03ceea8d4 100644 --- a/plugins/core-claude-light/workflows/research-flow.md +++ b/plugins/core-claude-light/workflows/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-claude/skills/research/README.md b/plugins/core-claude/skills/research/README.md index 65221f7da..5e9af8f6b 100644 --- a/plugins/core-claude/skills/research/README.md +++ b/plugins/core-claude/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `workflows/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-claude/skills/research/SKILL.md b/plugins/core-claude/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-claude/skills/research/SKILL.md +++ b/plugins/core-claude/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-claude/workflows/modernization-flow-analysis.md b/plugins/core-claude/workflows/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-claude/workflows/modernization-flow-analysis.md +++ b/plugins/core-claude/workflows/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-claude/workflows/modernization-flow-implement.md b/plugins/core-claude/workflows/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-claude/workflows/modernization-flow-implement.md +++ b/plugins/core-claude/workflows/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-claude/workflows/modernization-flow-mapping.md b/plugins/core-claude/workflows/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-claude/workflows/modernization-flow-mapping.md +++ b/plugins/core-claude/workflows/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-claude/workflows/modernization-flow.md b/plugins/core-claude/workflows/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-claude/workflows/modernization-flow.md +++ b/plugins/core-claude/workflows/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-claude/workflows/research-flow.md b/plugins/core-claude/workflows/research-flow.md index 2bc2b4816..03ceea8d4 100644 --- a/plugins/core-claude/workflows/research-flow.md +++ b/plugins/core-claude/workflows/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md b/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md index fa1dbfdd8..498f7079d 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/SKILL.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md index 2b0a8ccb9..6f1dd855b 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -17,7 +17,7 @@ ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-implement.md b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-implement.md index 55a50e2e7..d525ae744 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-implement.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-implement.md @@ -3,7 +3,7 @@ ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md index ff77204ed..aa7f728ad 100644 --- a/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-codex-light/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-codex-light/.agents/skills/research-flow/SKILL.md b/plugins/core-codex-light/.agents/skills/research-flow/SKILL.md index 20db94499..87a5ac82f 100644 --- a/plugins/core-codex-light/.agents/skills/research-flow/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/research-flow/SKILL.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-codex-light/.agents/skills/research/README.md b/plugins/core-codex-light/.agents/skills/research/README.md index b036c061f..2dac9d281 100644 --- a/plugins/core-codex-light/.agents/skills/research/README.md +++ b/plugins/core-codex-light/.agents/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-codex-light/.agents/skills/research/SKILL.md b/plugins/core-codex-light/.agents/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-codex-light/.agents/skills/research/SKILL.md +++ b/plugins/core-codex-light/.agents/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md b/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md index fa1dbfdd8..498f7079d 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/SKILL.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md index 2b0a8ccb9..6f1dd855b 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-analysis.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -17,7 +17,7 @@ ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-implement.md b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-implement.md index 55a50e2e7..d525ae744 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-implement.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-implement.md @@ -3,7 +3,7 @@ ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md index ff77204ed..aa7f728ad 100644 --- a/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md +++ b/plugins/core-codex/.agents/skills/modernization-flow/phases/modernization-flow-mapping.md @@ -3,7 +3,7 @@ ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -15,7 +15,7 @@ ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -44,7 +44,7 @@ ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-codex/.agents/skills/research-flow/SKILL.md b/plugins/core-codex/.agents/skills/research-flow/SKILL.md index 20db94499..87a5ac82f 100644 --- a/plugins/core-codex/.agents/skills/research-flow/SKILL.md +++ b/plugins/core-codex/.agents/skills/research-flow/SKILL.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-codex/.agents/skills/research/README.md b/plugins/core-codex/.agents/skills/research/README.md index b036c061f..2dac9d281 100644 --- a/plugins/core-codex/.agents/skills/research/README.md +++ b/plugins/core-codex/.agents/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `skills/self-help-flow/SKILL.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-codex/.agents/skills/research/SKILL.md b/plugins/core-codex/.agents/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-codex/.agents/skills/research/SKILL.md +++ b/plugins/core-codex/.agents/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-copilot-light/commands/modernization-flow-analysis.md b/plugins/core-copilot-light/commands/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-copilot-light/commands/modernization-flow-analysis.md +++ b/plugins/core-copilot-light/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-copilot-light/commands/modernization-flow-implement.md b/plugins/core-copilot-light/commands/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-copilot-light/commands/modernization-flow-implement.md +++ b/plugins/core-copilot-light/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-copilot-light/commands/modernization-flow-mapping.md b/plugins/core-copilot-light/commands/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-copilot-light/commands/modernization-flow-mapping.md +++ b/plugins/core-copilot-light/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-light/commands/modernization-flow.md b/plugins/core-copilot-light/commands/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-copilot-light/commands/modernization-flow.md +++ b/plugins/core-copilot-light/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-copilot-light/commands/research-flow.md b/plugins/core-copilot-light/commands/research-flow.md index 9776c6d6d..4d07fc32d 100644 --- a/plugins/core-copilot-light/commands/research-flow.md +++ b/plugins/core-copilot-light/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-copilot-light/skills/research/README.md b/plugins/core-copilot-light/skills/research/README.md index f4c58de1e..94d8c0bbe 100644 --- a/plugins/core-copilot-light/skills/research/README.md +++ b/plugins/core-copilot-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-copilot-light/skills/research/SKILL.md b/plugins/core-copilot-light/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-copilot-light/skills/research/SKILL.md +++ b/plugins/core-copilot-light/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-analysis.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-analysis.prompt.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-analysis.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-analysis.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-implement.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-implement.prompt.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-implement.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-implement.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow-mapping.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/modernization-flow.prompt.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-copilot-standalone-light/.github/prompts/research-flow.prompt.md b/plugins/core-copilot-standalone-light/.github/prompts/research-flow.prompt.md index 9776c6d6d..4d07fc32d 100644 --- a/plugins/core-copilot-standalone-light/.github/prompts/research-flow.prompt.md +++ b/plugins/core-copilot-standalone-light/.github/prompts/research-flow.prompt.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-copilot-standalone-light/.github/skills/research/README.md b/plugins/core-copilot-standalone-light/.github/skills/research/README.md index 7af890869..f799337ca 100644 --- a/plugins/core-copilot-standalone-light/.github/skills/research/README.md +++ b/plugins/core-copilot-standalone-light/.github/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `prompts/self-help-flow.prompt.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md b/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md +++ b/plugins/core-copilot-standalone-light/.github/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-analysis.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-analysis.prompt.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-analysis.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-analysis.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-implement.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-implement.prompt.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-implement.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-implement.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow-mapping.prompt.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md b/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/modernization-flow.prompt.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-copilot-standalone/.github/prompts/research-flow.prompt.md b/plugins/core-copilot-standalone/.github/prompts/research-flow.prompt.md index 9776c6d6d..4d07fc32d 100644 --- a/plugins/core-copilot-standalone/.github/prompts/research-flow.prompt.md +++ b/plugins/core-copilot-standalone/.github/prompts/research-flow.prompt.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-copilot-standalone/.github/skills/research/README.md b/plugins/core-copilot-standalone/.github/skills/research/README.md index 7af890869..f799337ca 100644 --- a/plugins/core-copilot-standalone/.github/skills/research/README.md +++ b/plugins/core-copilot-standalone/.github/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `prompts/self-help-flow.prompt.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-copilot-standalone/.github/skills/research/SKILL.md b/plugins/core-copilot-standalone/.github/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-copilot-standalone/.github/skills/research/SKILL.md +++ b/plugins/core-copilot-standalone/.github/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-copilot/commands/modernization-flow-analysis.md b/plugins/core-copilot/commands/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-copilot/commands/modernization-flow-analysis.md +++ b/plugins/core-copilot/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-copilot/commands/modernization-flow-implement.md b/plugins/core-copilot/commands/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-copilot/commands/modernization-flow-implement.md +++ b/plugins/core-copilot/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-copilot/commands/modernization-flow-mapping.md b/plugins/core-copilot/commands/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-copilot/commands/modernization-flow-mapping.md +++ b/plugins/core-copilot/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-copilot/commands/modernization-flow.md b/plugins/core-copilot/commands/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-copilot/commands/modernization-flow.md +++ b/plugins/core-copilot/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-copilot/commands/research-flow.md b/plugins/core-copilot/commands/research-flow.md index 9776c6d6d..4d07fc32d 100644 --- a/plugins/core-copilot/commands/research-flow.md +++ b/plugins/core-copilot/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-copilot/skills/research/README.md b/plugins/core-copilot/skills/research/README.md index f4c58de1e..94d8c0bbe 100644 --- a/plugins/core-copilot/skills/research/README.md +++ b/plugins/core-copilot/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-copilot/skills/research/SKILL.md b/plugins/core-copilot/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-copilot/skills/research/SKILL.md +++ b/plugins/core-copilot/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-cursor-light/commands/modernization-flow-analysis.md b/plugins/core-cursor-light/commands/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-cursor-light/commands/modernization-flow-analysis.md +++ b/plugins/core-cursor-light/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-cursor-light/commands/modernization-flow-implement.md b/plugins/core-cursor-light/commands/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-cursor-light/commands/modernization-flow-implement.md +++ b/plugins/core-cursor-light/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-cursor-light/commands/modernization-flow-mapping.md b/plugins/core-cursor-light/commands/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-cursor-light/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-light/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-light/commands/modernization-flow.md b/plugins/core-cursor-light/commands/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-cursor-light/commands/modernization-flow.md +++ b/plugins/core-cursor-light/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-cursor-light/commands/research-flow.md b/plugins/core-cursor-light/commands/research-flow.md index 8f0f040cf..bdace9f9b 100644 --- a/plugins/core-cursor-light/commands/research-flow.md +++ b/plugins/core-cursor-light/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-cursor-light/skills/research/README.md b/plugins/core-cursor-light/skills/research/README.md index 942562b7f..b90da6f5e 100644 --- a/plugins/core-cursor-light/skills/research/README.md +++ b/plugins/core-cursor-light/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-cursor-light/skills/research/SKILL.md b/plugins/core-cursor-light/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-cursor-light/skills/research/SKILL.md +++ b/plugins/core-cursor-light/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-analysis.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-analysis.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-implement.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-implement.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-cursor-standalone-light/.cursor/commands/research-flow.md b/plugins/core-cursor-standalone-light/.cursor/commands/research-flow.md index 8f0f040cf..bdace9f9b 100644 --- a/plugins/core-cursor-standalone-light/.cursor/commands/research-flow.md +++ b/plugins/core-cursor-standalone-light/.cursor/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md b/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md index 942562b7f..b90da6f5e 100644 --- a/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md +++ b/plugins/core-cursor-standalone-light/.cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md b/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md +++ b/plugins/core-cursor-standalone-light/.cursor/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-analysis.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-analysis.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-implement.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-implement.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md +++ b/plugins/core-cursor-standalone/.cursor/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-cursor-standalone/.cursor/commands/research-flow.md b/plugins/core-cursor-standalone/.cursor/commands/research-flow.md index 8f0f040cf..bdace9f9b 100644 --- a/plugins/core-cursor-standalone/.cursor/commands/research-flow.md +++ b/plugins/core-cursor-standalone/.cursor/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-cursor-standalone/.cursor/skills/research/README.md b/plugins/core-cursor-standalone/.cursor/skills/research/README.md index 942562b7f..b90da6f5e 100644 --- a/plugins/core-cursor-standalone/.cursor/skills/research/README.md +++ b/plugins/core-cursor-standalone/.cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md b/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md +++ b/plugins/core-cursor-standalone/.cursor/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way) diff --git a/plugins/core-cursor/commands/modernization-flow-analysis.md b/plugins/core-cursor/commands/modernization-flow-analysis.md index 65ed95af8..d142a80bc 100644 --- a/plugins/core-cursor/commands/modernization-flow-analysis.md +++ b/plugins/core-cursor/commands/modernization-flow-analysis.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 2: Old Code Analysis** -1. Complete Phase 1 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 1 first, unless on the reduced route recorded in the state file 2. Process each project separately 3. Analyze code and tests 4. Create original-code-specs files @@ -26,7 +26,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 1 (all reference code specs created). EXCEPTION — on the reduced route recorded in the state file (small in-place modernization, no reusable reference-library analysis needed), Phase 1 is skipped: proceed with no reference specs and note this explicitly in the original-code-specs files. +**PREREQUISITES:** Complete Phase 1 (all reference code specs created), unless on the reduced route recorded in the state file **OBJECTIVES:** - Create complete, factual specification of existing code diff --git a/plugins/core-cursor/commands/modernization-flow-implement.md b/plugins/core-cursor/commands/modernization-flow-implement.md index 57460b63d..2c6391d04 100644 --- a/plugins/core-cursor/commands/modernization-flow-implement.md +++ b/plugins/core-cursor/commands/modernization-flow-implement.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Context **PREREQUISITES:** -- Complete Phase 7 (final cross-project review). EXCEPTION — on the reduced route recorded in the state file, Phase 7 is skipped: complete Phase 6 instead; the single-project scope note in the state file replaces the cross-project review. +- Complete Phase 7 (final cross-project review), unless on the reduced route recorded in the state file - **EXPLICIT HUMAN APPROVAL** of all target specs ## Requirements diff --git a/plugins/core-cursor/commands/modernization-flow-mapping.md b/plugins/core-cursor/commands/modernization-flow-mapping.md index 38b84b8e4..43b7bd1a4 100644 --- a/plugins/core-cursor/commands/modernization-flow-mapping.md +++ b/plugins/core-cursor/commands/modernization-flow-mapping.md @@ -12,7 +12,7 @@ baseSchema: docs/schemas/phase.md ## Phase Flow **Phase 6: Implementation Mapping** -1. Complete Phase 5 first (skipped ONLY on the reduced route recorded in the state file) +1. Complete Phase 5 first, unless on the reduced route recorded in the state file 2. Define technical approaches 3. Compare multiple alternatives 4. Map source to target @@ -24,7 +24,7 @@ baseSchema: docs/schemas/phase.md ## Context -**PREREQUISITES:** Complete Phase 5 (cross-project analysis). EXCEPTION — on the reduced route recorded in the state file (single-project mapping scope), Phase 5 is skipped: `docs/cross-project-analysis.md` does not exist; use the single-project scope note in the state file's `Reduced Route` section instead. +**PREREQUISITES:** Complete Phase 5 (cross-project analysis), unless on the reduced route recorded in the state file **OBJECTIVES:** - Define technical approaches for modernization @@ -53,7 +53,7 @@ baseSchema: docs/schemas/phase.md ## Requirements - **MUST** process each project separately -- **MUST** re-read `docs/cross-project-analysis.md` (on the reduced route: the single-project scope note in the state file instead) +- **MUST** re-read `docs/cross-project-analysis.md` if it exists - **MUST** re-read matching `docs/original-code-specs-.md` - **MUST** re-read both incoming/outgoing dependency if those exist `docs/target-code-specs-.md` - **MUST** create `docs/target-code-specs-.md` for each project diff --git a/plugins/core-cursor/commands/modernization-flow.md b/plugins/core-cursor/commands/modernization-flow.md index 814f31ec0..95dcd72a2 100644 --- a/plugins/core-cursor/commands/modernization-flow.md +++ b/plugins/core-cursor/commands/modernization-flow.md @@ -14,10 +14,10 @@ This modernization process is designed to systematically analyze, specify, and i ### Critical Requirements - **ONE PHASE AT A TIME**: Read one phase file, execute it completely, update state file, then move to next phase. -- **DO NOT SKIP PHASES**: Each phase builds on the previous one. ONLY exception: the reduced route (see User Customizations). +- **DO NOT SKIP PHASES**: Each phase builds on the previous one. - **STATE TRACKING**: After each phase, update `agents/modernization-flow-state.md` with completion status, keep this document concise. - **USER CONFIRMATION**: Wait for user confirmation before proceeding to next phase. -- Make sure to have todo tasks for each phase! Do not skip phases (except on the reduced route)! +- Make sure to have todo tasks for each phase! Do not skip phases! - **REMEMBER** Do not put decisions/recommendations/suggestions/implementation code for the next phases in the output documents! The decisions are made at phase 6 and review at phase 7 AFTER full analysis. Respect that. - **MUST** Create and update files that ARE EXPLICITLY stated here, ASK for permission to create any other file! - **MUST** use grep/search on large files to read "table of contents" and then use line ranges to read what is needed. @@ -30,7 +30,7 @@ This modernization process is designed to systematically analyze, specify, and i - If user did not specify any preferences perform all steps except optional. - User CAN customize and ask only for specific steps OR steps could have been done already OR towards specific goal OR for specific case, in this case LISTEN and ADOPT to the user. -- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, unless you discover dependencies and project becomes bigger. This reduced route is an explicit exception to the phase-prerequisite contract: Phase 2 runs without Phase 1 ONLY after recording in the state file that no reusable reference-library analysis is needed (no prior reference specs exist and none is requested), and Phase 6 runs without Phase 5 ONLY under a bounded single-project mapping scope — cross-project findings are replaced by an explicit single-project scope note, written by the orchestrator in the state file's `Reduced Route` section before Phase 6. On the reduced route Phase 7 (Final Review) is skipped: it reviews cross-project artifacts that do not exist; the Phase 6 single-project note is its substitute. Record both decisions in the state file so the exception is auditable. +- Modernization scope could be small or could be done in place (example: upgrade Java 8 to 17 or 17 to 21), in this case only use `Phase 2` and `Phase 6`, skip the other phases and their prerequisites, record this in the state file, unless you discover dependencies and project becomes bigger. - If there is no need to rewrite code (example: upgrade .NET 6 to .NET 8), then take original and target code specs LIGHTLY, only document WHAT IS NECESSARY TO CHANGE (example: new way to use app builder). ## Modernization Flow - Phase Overview @@ -114,10 +114,6 @@ Create/update `agents/modernization-flow-state.md` briefly after each phase usin - [ ] Phase 7: Final Review - Not Started - [ ] Implementation Phase - Not Started -## Reduced Route - -Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because [no reusable reference-library analysis needed: …]; Phase 5 skipped because [single-project mapping scope: …]; Single-project scope note: [project, why no cross-project dependencies, boundaries]; Phase 7 skipped (substituted by the Phase 6 single-project note). - ## Phase Details ### Phase 1 @@ -131,7 +127,7 @@ Reduced route: no — full sequence. If reduced: yes — Phase 1 skipped because ## Important Notes -- **Sequential Execution Required**: Phases build on each other and must be done in order — the ONLY exception is the reduced route recorded in the state file (see User Customizations), whose skipped-phase substitutes are defined in the affected phase files. +- **Sequential Execution Required**: Phases build on each other and must be done in order - **No Assumptions**: If information is missing, track it in spec files - **User Guidance**: Ask questions to resolve critical unknowns - **Evidence-Based**: All findings must reference actual code diff --git a/plugins/core-cursor/commands/research-flow.md b/plugins/core-cursor/commands/research-flow.md index 8f0f040cf..bdace9f9b 100644 --- a/plugins/core-cursor/commands/research-flow.md +++ b/plugins/core-cursor/commands/research-flow.md @@ -45,7 +45,7 @@ Orchestrates deep research via meta-prompting: craft an optimized research promp -1. Execute the approved research prompt directly in this researcher context (no nested subagent; the workflow owns dispatch). +1. Execute the approved research prompt as a separate subagent. 2. Input: approved `research-prompt.md`. Output: `docs/-research.md`. 3. Required skills: `research` 4. Update `research-flow-state.md`. diff --git a/plugins/core-cursor/skills/research/README.md b/plugins/core-cursor/skills/research/README.md index 942562b7f..b90da6f5e 100644 --- a/plugins/core-cursor/skills/research/README.md +++ b/plugins/core-cursor/skills/research/README.md @@ -1,5 +1,5 @@ # research -Meta-prompting skill: craft an optimized research prompt first, then execute that prompt (never ad-hoc research). Execution is caller-scoped: a spawned researcher runs it in its own context and returns parallel-slice proposals to its parent; a top-level agent dispatches it to `researcher` subagents under orchestration. Nested dispatch stays prohibited. +Meta-prompting skill: craft an optimized research prompt first, then execute it as a separate subagent — never research directly. Researches any question grounded in your project, with real sources. Compares options and validates its own conclusions. @@ -10,14 +10,14 @@ Fixes the failure mode where an AI treats its training-data knowledge as suffici Actor: the `researcher` subagent (`mode: subagent`, `readonly: false`). Trigger: systematic research needing grounded references, multi-option analysis, and self-validation; skip for simple lookups or single-source questions. Prerequisite: "All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed." Not meant for casual direct invocation — `commands/self-help-flow.md` flags `/research ...` as a BAD example because "using the skill directly bypasses the structured research workflow; PRIORITY RULE applies," steering real usage through `research-flow` instead. ## How it works -Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, caller-scoped parallel-work handling (spawned researcher proposes slices to its parent; top-level agent dispatches them)) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. +Single flat SKILL.md, no `assets/` or `references/`. Flow: `` (senior research specialist, meta-prompting) → `` gate → `` (prep-step prerequisite, meta-prompting approach, a hard restriction against touching CONTEXT.md/ARCHITECTURE.md/IMPLEMENTATION.md) → ``, itself split in two: "Research rules" govern how research is actually carried out (plan, tree-of-thoughts with ≥3 options, ongoing `research-state.md`, save to `docs/-research.md`, grounding/HITL discipline, parallel subagents) and "Enforcement rules for the generated research prompt" are 5 MUSTs that this skill bakes into the prompt it hands off, not into its own execution. Invoked by the `researcher` subagent inside `research-flow` phase 3 (`execute_research`), and ad hoc by orchestration's team-manager asset for external-knowledge lookups. ## Mental hooks & unexpected rules -- "craft an optimized research prompt first, then execute that prompt — never ad-hoc research" — this skill's real output is a prompt, not an answer; answering research questions inline without the prepared prompt is out of contract. +- "craft an optimized research prompt first, then execute it — never research directly" — this skill's real output is a prompt, not an answer; answering research questions inline is out of contract. - "MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly" — research output is deliberately walled off from the project's core docs. - "Follow tree-of-thoughts pattern and analyze at least 3 options" — a single correct-looking option still fails the skill's contract. - "Fall back to anecdotal references, but call this out EXPLICITLY!" — anecdote is permitted only with a loud disclaimer, never silently blended with reputable sources. -- "MUST use authoritative sources available to the current environment" — a capability-aware source mandate written into the generated research prompt: DeepWiki and Context7 preferred when available, explicit fallbacks otherwise. +- "MUST use DeepWiki and Context7 when available" — a named-tool mandate written into the generated research prompt, conditional on the tools being installed. - "MUST think about and align consequences and consequences of consequences to prevent oversight" — second-order-effects analysis is required of the crafted prompt, not optional depth. ## Invariants — do not change diff --git a/plugins/core-cursor/skills/research/SKILL.md b/plugins/core-cursor/skills/research/SKILL.md index eefd47ae6..bd2376e68 100644 --- a/plugins/core-cursor/skills/research/SKILL.md +++ b/plugins/core-cursor/skills/research/SKILL.md @@ -10,7 +10,7 @@ baseSchema: docs/schemas/skill.md -You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute that prompt — never ad-hoc research. +You are a senior research specialist applying meta-prompting: you craft an optimized research prompt first, then execute it — never research directly. @@ -21,7 +21,7 @@ Systematic research: grounded references, multiple-options analysis, self-valida - All Rosetta prep steps MUST be FULLY completed, load-project-context skill loaded and fully executed -- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it per the caller: when invoked as a spawned researcher, execute the prepared prompt directly in this context (subagents cannot dispatch subagents); when invoked directly by a top-level agent, the prepared prompt MAY be dispatched to a `researcher` subagent under orchestration (the top-level agent owns dispatch). +- Meta-prompting approach: prepare an optimized research prompt enforcing all rules below, then execute it as a separate subagent - MUST NOT update CONTEXT.md, ARCHITECTURE.md, IMPLEMENTATION.md, and create any other documents EXCEPT those mentioned explicitly @@ -44,13 +44,13 @@ Research rules: - Search documentation for libraries, versions, and issues not in built-in knowledge - MUST be cautious of LLM context: use grep, search, and similar techniques and tools - Ask user questions during research to resolve unknowns and validate direction -- When individual ideas or areas would benefit from parallel exploration: as a spawned researcher, propose bounded parallel slices (scope, dependencies, expected evidence, write-collision risk) back to the parent — do NOT spawn subagents from here (subagents cannot dispatch subagents); as a top-level agent, dispatch those slices to `researcher` subagents under orchestration. +- Spawn parallel subagents to go over individual ideas or areas - Use synthesis and comparison approach Enforcement rules for the generated research prompt: 1. MUST use todo tasks -2. MUST use authoritative sources available to the current environment: prefer DeepWiki and Context7 when those integrations are available; otherwise use the project's own documentation, vendor docs, and web search as fallback. Name the source actually used for every grounded claim. +2. MUST use DeepWiki and Context7 when available 3. MUST create and update state md file after each task 4. MUST output result file section by section as soon as each section becomes available 5. MUST think about and align consequences and consequences of consequences to prevent oversight (example: doing X leads to Y, which affects Z, thus it should be done ABC way)