You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Oct 3, 2026. It is now read-only.
The owner subsequently requested review and merge. All four PRs were reviewed and squash-merged in dependency order, with an exact-head SHA guard and green Verify CI before each merge:
Normal integration commits preserved all exact branch deployment guards; no branch history was rewritten and no worktree was deleted or reset. Affected ROADMAP status now reflects the merged recovery and Club behavior. Final integration CI passed 30 Node tests, 36 Chromium checks, typecheck, lint, build and HTTP/search smoke: https://github.com/WalksWithASwagger/wedges/actions/runs/34051376724 . Final main CI also passed: https://github.com/WalksWithASwagger/wedges/actions/runs/34051463010 . Staged integration diffs passed Gitleaks. All existing worktrees are clean; the original local main checkout remains untouched.
GitHub reports successful Production deployments for all four merge commits, including deployment 6296395710 for final main. Issues #13–#16 closed automatically. #18 is now ready after dependency review and issue lint; no new implementation lane was launched. Existing advisories, provider spending uncertainty, human pilot/AT coverage and paid visual semantic evaluation remain unresolved; these merges do not clear broader promotion gates. Native auto-merge remains disabled.
The section below is the historical handoff before the owner's merge authorization, not current delivery status.
Historical pre-merge swarm handoff — 2026-09-06
The requested run activated #13, #14 and #15 in that order, followed by the coordinator's scoped #16 report. Each used an isolated codex/ worktree from dc75efe; existing checkouts and uncommitted work were preserved. No native runner/global configuration was installed.
Stops generated member endorsement; preserves legacy data and failed drafts. 20 Node tests, 18 Chromium checks, build, lint/typecheck, MCP smoke, 30 search checks; desktop/mobile inspected. Final docs-only correction has green CI too.
Bounded offline visual experiment. 29 Node tests, lint/typecheck and offline preflight; independent review passed. No paid or human semantic evaluation.
Release-evidence report; unchanged 13-advisory baseline, firewall/variable-name observations and unverified provider-spend gate. 19 tests, lint/typecheck, link and secret checks; independent review passed.
All four exact PR heads passed Verify CI and are individually mergeable against current main. Local worktrees are clean. No new PR was merged and native auto-merge remains disabled. review-ready means ready for human review, not permission to deploy or proof of product/security quality. No live paid evals, participant outreach or production configuration changes occurred.
Delivery side effect: the first #14 push automatically created a non-production Vercel preview at 165789c before the guard was added. It was already READY when cancellation eligibility was checked; it was not cancelled or deleted. No production alias was changed. Subsequent heads carry an exact-branch no-deployment guard; #13/#15/#16 were guarded before publishing.
Integration ownership: all four PRs add a different exact branch entry to vercel.json. Their entries must be combined rather than discarded when merging sequentially. #13/#14 edit distinct README paragraphs; #14/#16 edit distinct roadmap areas. Revalidate the final combined docs and CI. This run did not blindly rebase, reset or merge the branches.
#17–#22 remain blocked on their existing human/dependency gates. A green offline experiment does not unlock #22; a successful report does not establish a provider spending cap. Review these PRs before starting another implementation wave.
Context
Direction: Wedges helps a creative author bring a piece of work and relevant taste, inspect cited interpretation, make their own decisions, keep the evidence, and return with sharper context. Preserve its ink/paper/blood visual identity, direct voice, portable artifacts and human judgment. The proposed benefits below are hypotheses until their evidence gates pass.
Completed: #9/#10 introduced bounded cited solo critique; #11/#12 added browser Review → Decide → Export. PR #12 merged on 2026-09-06 as dc75efe5ce87ffe3583657f91bda86b96bb4a812. Its exact-head CI and the subsequent main CI both passed. Fresh local revalidation passed 19 tests, typecheck and lint. No other PR was open after the merge.
This tracker coordinates ten bounded packets; it is not an implementation issue or authorization to launch a swarm. Issue publication was requested; future agents must be assigned a specific packet. No automatic merge or deployment is enabled by these issues.
Concrete public-review release decision with runtime/advisory/spending evidence
Release-readiness doc; one small ROADMAP status update
All four passed issue-shape lint before dispatch; their current review state is recorded above. Recommended first three implementation/evidence lanes: #13, #14, #15. #16 can investigate read-only alongside them, but its ROADMAP edit must wait for #14's ownership to clear. Do not let two agents edit the same shared file just because their issue labels are ready.
Still-image review with referenced assets and portable author decisions
#13 merged, #17 positive/narrowed, and #15 live semantic/payload gates passed
At initial publication these six carried blocked; #18 is now agent:ready after #24 merged. #17/#19/#21/#22 also carry needs-human. A completed protocol, synthetic fixture pass or closed experiment issue does not satisfy its human/semantic gate.
Each implementation outcome lands in one focused reviewed PR linked to its packet, with acceptance evidence and actual checks.
Clear blocked/needs-human only after linked PRs and named evidence/decisions are reviewed; then re-lint and apply agent:ready. Do not run stop-labeled issues.
Preserve text review v1 and existing profile consumers; any new artifact/migration has compatibility tests and explicit data-preservation/rollback limits.
Human decisions remain authored, AI interpretation remains attributed, and no automatic approval, revision or taste update appears.
Keep old docs/TODOs as evidence; reconcile affected claims in the delivering PR rather than treating the old roadmap as a mandatory feature list.
Tests/Evals and Verification
Every packet contains exact owned surfaces, acceptance criteria, exclusions, dependencies, risks, commands and UI/eval requirements. Revalidate the baseline on each branch. Existing deterministic checks cover typecheck, Node tests, lint, production build, Chromium, MCP smoke and search smoke; future model behavior changes need their own bounded semantic evidence. Never weaken checks or call unavailable checks passing.
The ten packets passed the existing shared issue linter using its canonical section/checklist gate. Ready candidates also pass with their actual labels; stop-labeled packets intentionally do not pass the runnable gate. Wedges currently has no repository-local agentic/contract.json, scripts/agentic/issue_lint.py or registered runner contract. This validates issue structure, not automatic runner enrollment. Do not copy KB configuration, install a swarm framework, create a Linear mirror or modify shared/global skills to make labels appear operational. A coordinator can assign these bounded packets directly.
Optional experiments — parked until a concrete trigger
These remain visible here instead of becoming unconditional feature work:
Explicit AI assistance in Film Club: revisit only after Let real members comment on a selected Film Club work snapshot #20's human workflow is used and authors request it. One user-requested invocation, a profile owner's explicit permission for lens use, clear AI attribution, bounded cost and no member endorsement. Reuse cited critique primitives where valid. A future focused packet must test attribution, permission, failures and value; no automatic fan-out or room-average summary.
Model comparison: if pilot/eval evidence reveals a quality or latency problem, verify currently shipping provider models and installed SDK options from official docs, then run a capped, blind comparison against the existing model on the same fixtures. Compare semantic fit, author agency, grounded evidence, p50/p95 latency and actual cost. No default switch based on novelty or a model name from an old plan.
MCP Apps presentation: if observed context-switching is a real obstacle, test reuse of the review UI in one currently supported real host with an explicit compatibility matrix. No custom agent host, universal host-support claim or framework dependency until the experiment earns it.
Optional exercises: if Promote explicit author decisions into scoped, portable taste notes #21 users lack useful source taste, test a contextual link into an existing Mirror/Taste Audit/Selector exercise. Do not require a questionnaire or delete existing exercises without checking MCP/prompt/profile consumers.
Create an implementation issue for one of these only when the trigger, scope, owner and success criterion are established. A parked idea is not an instruction for an autonomous agent.
Agent Instructions
Before every lane, inspect current instructions, Git status, all worktrees, relevant live PRs and dependencies. Use an isolated codex/ branch/worktree and preserve others' work. Main in the existing checkout was intentionally left behind; branch from the verified remote state without overwriting or syncing another checkout. Never stash/reset/rebase/delete others' state or force-push. Future PRs require explicit review; this tracker grants no merge/deploy permission.
No participant outreach, private-work publication, production migration, secret changes or unbounded paid evaluations. Stop at each named gate and leave evidence plus an exact next step.
Out of Scope / deferred maintenance
No agent conductor, automatic publishing, generic chatbot, full video editor, accounts/cloud sync, embeddings/fine-tuning, taste commons/averages, votes, mandatory onboarding, repository-wide cleanup or new global agent infrastructure. Older registry/discoverability/analytics/cross-site-link work remains outside this transformation unless separately prioritized. Do not clean up dormant worktrees or delete old profile/scoring fields based on age.
Recommended next action
Assign #18 (atomic Club room mutations) as the next focused development lane. Its #14 dependency is satisfied by merged PR #24; the updated packet passed the shared issue-shape gate with agent:ready,bug. Real Redis integration evidence remains required. No next implementation wave was started during merge review. #17 and #19–#22 retain their existing human/dependency gates.
Wedges: portable creative review — delivery queue
Authorized merge completion — 2026-09-06
The owner subsequently requested review and merge. All four PRs were reviewed and squash-merged in dependency order, with an exact-head SHA guard and green Verify CI before each merge:
c524c9eae8b946e792e60ef7ab39cc34bf1096b8aa4e1e76b6ffa04115d7013bda79245ec8b4f42d8b740964ba5d62db221598cd8fbc526d028c68b4bd6fec810f5c18b582d76ab302212913eef30d06Normal integration commits preserved all exact branch deployment guards; no branch history was rewritten and no worktree was deleted or reset. Affected ROADMAP status now reflects the merged recovery and Club behavior. Final integration CI passed 30 Node tests, 36 Chromium checks, typecheck, lint, build and HTTP/search smoke: https://github.com/WalksWithASwagger/wedges/actions/runs/34051376724 . Final main CI also passed: https://github.com/WalksWithASwagger/wedges/actions/runs/34051463010 . Staged integration diffs passed Gitleaks. All existing worktrees are clean; the original local main checkout remains untouched.
GitHub reports successful Production deployments for all four merge commits, including deployment 6296395710 for final main. Issues #13–#16 closed automatically. #18 is now ready after dependency review and issue lint; no new implementation lane was launched. Existing advisories, provider spending uncertainty, human pilot/AT coverage and paid visual semantic evaluation remain unresolved; these merges do not clear broader promotion gates. Native auto-merge remains disabled.
The section below is the historical handoff before the owner's merge authorization, not current delivery status.
Historical pre-merge swarm handoff — 2026-09-06
The requested run activated #13, #14 and #15 in that order, followed by the coordinator's scoped #16 report. Each used an isolated codex/ worktree from dc75efe; existing checkouts and uncommitted work were preserved. No native runner/global configuration was installed.
All four exact PR heads passed Verify CI and are individually mergeable against current main. Local worktrees are clean. No new PR was merged and native auto-merge remains disabled.
review-readymeans ready for human review, not permission to deploy or proof of product/security quality. No live paid evals, participant outreach or production configuration changes occurred.Delivery side effect: the first #14 push automatically created a non-production Vercel preview at 165789c before the guard was added. It was already READY when cancellation eligibility was checked; it was not cancelled or deleted. No production alias was changed. Subsequent heads carry an exact-branch no-deployment guard; #13/#15/#16 were guarded before publishing.
Integration ownership: all four PRs add a different exact branch entry to vercel.json. Their entries must be combined rather than discarded when merging sequentially. #13/#14 edit distinct README paragraphs; #14/#16 edit distinct roadmap areas. Revalidate the final combined docs and CI. This run did not blindly rebase, reset or merge the branches.
#17–#22 remain blocked on their existing human/dependency gates. A green offline experiment does not unlock #22; a successful report does not establish a provider spending cap. Review these PRs before starting another implementation wave.
Context
Direction: Wedges helps a creative author bring a piece of work and relevant taste, inspect cited interpretation, make their own decisions, keep the evidence, and return with sharper context. Preserve its ink/paper/blood visual identity, direct voice, portable artifacts and human judgment. The proposed benefits below are hypotheses until their evidence gates pass.
Completed: #9/#10 introduced bounded cited solo critique; #11/#12 added browser Review → Decide → Export. PR #12 merged on 2026-09-06 as
dc75efe5ce87ffe3583657f91bda86b96bb4a812. Its exact-head CI and the subsequent main CI both passed. Fresh local revalidation passed 19 tests, typecheck and lint. No other PR was open after the merge.This tracker coordinates ten bounded packets; it is not an implementation issue or authorization to launch a swarm. Issue publication was requested; future agents must be assigned a specific packet. No automatic merge or deployment is enabled by these issues.
Initial dispatch plan and ownership
All four passed issue-shape lint before dispatch; their current review state is recorded above. Recommended first three implementation/evidence lanes: #13, #14, #15. #16 can investigate read-only alongside them, but its ROADMAP edit must wait for #14's ownership to clear. Do not let two agents edit the same shared file just because their issue labels are ready.
Sequential work and explicit gates
At initial publication these six carried
blocked; #18 is nowagent:readyafter #24 merged. #17/#19/#21/#22 also carryneeds-human. A completed protocol, synthetic fixture pass or closed experiment issue does not satisfy its human/semantic gate.Dependency graph:
Acceptance Criteria
blocked/needs-humanonly after linked PRs and named evidence/decisions are reviewed; then re-lint and applyagent:ready. Do not run stop-labeled issues.Tests/Evals and Verification
Every packet contains exact owned surfaces, acceptance criteria, exclusions, dependencies, risks, commands and UI/eval requirements. Revalidate the baseline on each branch. Existing deterministic checks cover typecheck, Node tests, lint, production build, Chromium, MCP smoke and search smoke; future model behavior changes need their own bounded semantic evidence. Never weaken checks or call unavailable checks passing.
The ten packets passed the existing shared issue linter using its canonical section/checklist gate. Ready candidates also pass with their actual labels; stop-labeled packets intentionally do not pass the runnable gate. Wedges currently has no repository-local
agentic/contract.json,scripts/agentic/issue_lint.pyor registered runner contract. This validates issue structure, not automatic runner enrollment. Do not copy KB configuration, install a swarm framework, create a Linear mirror or modify shared/global skills to make labels appear operational. A coordinator can assign these bounded packets directly.Optional experiments — parked until a concrete trigger
These remain visible here instead of becoming unconditional feature work:
Create an implementation issue for one of these only when the trigger, scope, owner and success criterion are established. A parked idea is not an instruction for an autonomous agent.
Agent Instructions
Before every lane, inspect current instructions, Git status, all worktrees, relevant live PRs and dependencies. Use an isolated
codex/branch/worktree and preserve others' work. Main in the existing checkout was intentionally left behind; branch from the verified remote state without overwriting or syncing another checkout. Never stash/reset/rebase/delete others' state or force-push. Future PRs require explicit review; this tracker grants no merge/deploy permission.No participant outreach, private-work publication, production migration, secret changes or unbounded paid evaluations. Stop at each named gate and leave evidence plus an exact next step.
Out of Scope / deferred maintenance
No agent conductor, automatic publishing, generic chatbot, full video editor, accounts/cloud sync, embeddings/fine-tuning, taste commons/averages, votes, mandatory onboarding, repository-wide cleanup or new global agent infrastructure. Older registry/discoverability/analytics/cross-site-link work remains outside this transformation unless separately prioritized. Do not clean up dormant worktrees or delete old profile/scoring fields based on age.
Recommended next action
Assign #18 (atomic Club room mutations) as the next focused development lane. Its #14 dependency is satisfied by merged PR #24; the updated packet passed the shared issue-shape gate with
agent:ready,bug. Real Redis integration evidence remains required. No next implementation wave was started during merge review. #17 and #19–#22 retain their existing human/dependency gates.