Repository navigation
Description
Confirmed chat-access modes — 2026-10-12 (Asia/Tokyo)
The Human confirmed that ordinary users need not manage pairing/permissions merely to talk to a Bot. Describe the existing optional sender policy and conversation exceptions as chat access, not as a security bypass:
| Mode | Ordinary chat eligibility | Management authority |
|---|---|---|
| Open chat | With Bot-scoped restricted chat off, otherwise permitted DMs/@mentions do not require ordinary user pairing. | No new management capability, Role, delegation or machine Owner right. |
| Paired users only | Explicit restricted chat requires current reviewed ordinary pairing for intentional requests in supported scopes. | Management checks remain independent and current. |
| Paired users only + selected open conversations | Keep restricted chat on; explicitly allow unpaired visitors in an exact permitted external conversation. | The exception grants request initiation only, not approval/rule saving or administrator enrollment. |
The exact exception is scoped to PersonaBot + current App Binding/fingerprint + authenticated workspace/conversation ID, not a channel name or shared local Channel. Platform reach, app/Bot readiness, durable Block, current explicit Deny semantics, reception, mention/wake rules and bounds remain independent gates. “Open” means pairing-free ordinary chat, not unlimited messages, all-message collection, guaranteed model replies or arbitrary-resource/private-data isolation. Preserve the open default and explicit opt-in; do not silently apply new restrictions to existing traffic.
Current delivery and next tracer
- Lark tracer: explicit open groups and non-triggering passive context #1376 remains closed/accepted for Lark exact-group visitor access and passive context. Do not reopen it or reinterpret it as cross-platform proof.
- Slack tracer: ordinary roles and reviewed chat pairing with current-authority intake #1415 remains closed/accepted for Slack ordinary private-DM pairing. It does not qualify restricted Slack group requests or an open-channel exception.
- Slack tracer: explicit open-channel visitor chat under optional Bot pairing #1463 — Slack tracer: explicit open-channel visitor chat under optional Bot pairing is the new native child in Project M3: Bot 模式 IA 与聊天外壳(列表/Section/DM/本地消息) #10 / v1.3, ready-for-agent, Todo, with no implementation claim. It reuses the existing owner/policy/exception and delivers one public-channel restriction → explicit opening → fresh visitor @mention/original reply → close/refuse/restart flow. Lark tracer: explicit open groups and non-triggering passive context #1376 and Slack tracer: ordinary roles and reviewed chat pairing with current-authority intake #1415 are fulfilled native baseline dependencies.
- The present Bot-level pairing UI promises Lark conversations and Slack DMs; extending it to Slack groups requires explicit scope copy and compatibility/opt-in handling in Slack tracer: explicit open-channel visitor chat under optional Bot pairing #1463. Do not claim all-platform parity or silently widen existing restricted policies. The three rows above are product semantics; a new broad three-option UI is not required by this tracer.
- Slack tracer: verified pairing and native Allow once / Reject #1046 keeps Slack management pairing and native Allow once/Reject; Slack tracer: formal questions answered through qualified native controls #1047/Slack tracer: explicitly confirmed exact and scoped approval rules #1048/Slack group tracer: named management destinations with qualified decision actors #1051, Slack feedback tracer: received and answered reactions from canonical facts #1161 and existing owners retain their own scopes and gates. This update does not implement native management actions or change their dependencies.
The three-mode direction is confirmed; platform qualification remains evidence-bound. #1463 must provide focused production-seam regression, real authorized Slack evidence and GitHub-rendered UI proof, then pause at the first PR for Human QA. No implementation, receiver takeover, runtime access change, merge, package publication or deployment is authorized by this tracker update. Earlier dated coordination/acceptance sections below remain history; this section governs the chat-access clarification and new Slack tracer only.
Current extension delivery — Slack ordinary DM merged 2026-10-11
#1415 is closed/Done after PR #1436, merge 5c9614568c60d44a0823cf59afb7a1fa7f8fa0b3. The Human accepted the qualified native QA2 ordinary DM flow and authorized merge. Final acceptance records actual Web Role approval, fixed re-ask, new same-App replies/current permissions/three scoped directories and revoke/restart refusal, matched to canonical Source Event/Admission/Outbox.
Full reviewed-head CI and Standards/Spec reviews passed; the PR contains the full real workflow, light/dark/narrow revision comparisons and exact-head browser-reviewed HTML. Shared Slack QA1/QA2/QA3 setup remains machine-local alongside Lark, and the bilingual website guide now explains DM scopes, reinstallation and illustrated pairing. QA2 receiver and receiver-free screenshot Host are stopped; no production deployment or Provider pin promotion.
This acceptance covers ordinary Slack private chat only. #1046 retains native Slack management pairing/Allow once/Reject; #1404 retains later delegated pairing/code design, and all existing questions/rules/platform owners and blockers remain unchanged. Earlier dated pending #1415 summaries are superseded by this result; the Hub stays open for its remaining scope.
v1.3 release reconciliation — 2026-10-11
This Hub now tracks v1.3 cross-platform ordinary pairing/roles/group/person-authority parity, with Lark as the delivered first tracer. Preserve the native parent #1371 and closed children as historical specification/delivery; do not reopen accepted Lark slices to manufacture new work.
#1373–#1376 and #1392/#1406 are completed. PR #1407 is merged at 0c56a8b; Provider repair #20 is merged to its qualified line. The earlier ready-for-review/open-group summaries below are dated history. The first group tracer's task recorded completed; no remaining platform is inferred complete and no new owner is assigned.
| Remaining behavior | Owning issue / next gate |
|---|---|
| Slack ordinary DM review, primary Role, fresh-message intake, revoke/restart | New native child #1415, fulfilled baseline #1373/#1406; no claim yet |
| Slack native management pairing and Allow once/Reject | Existing #1046 under #1019; keep it separate from ordinary chat eligibility |
| Discord ordinary roles/groups and current permission queries | Remaining parity scope; coordinate #1042 and first current-product binding gate #1123 before deriving the next tracer |
| QQ and WeChat ordinary pairing/groups/person exceptions | Remaining parity scope; verify actual sender/conversation capabilities and equivalent secure interactions; do not assume QQ private DM or identical native controls |
| Multi-group union, explicit Deny, direct exceptions and actual platform management decisions | Reuse #1406's canonical resolver; first Slack approval integration follows #1046 and feedback from #1415 |
| Delegated Lark pairing review cards | Native child #1414 under #1404; task codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76 handed off for review; first real group-only flow accepted, independent Reject / delayed group-person Deny / restart / fresh 667 / revoke verified, no DM route; product #1418 + Provider BotHarness/dsh-im#21; QA1 stopped/released; merge pending |
| Reviewed one-time pairing codes | Existing #1404 retains codes as the explicit forwarding alternative after #1414 first-flow Human feedback; do not silently replace the newly confirmed card/delegation contract |
Natural-language Role policy remains behavior, not arbitrary resource ACL. Direct authenticated Web Human authority is unchanged. Group membership does not create chat eligibility, and pairing does not grant machine Owner/private-input rights.
Preserve original claims, native decision authorities and real-platform/Human-QA gates. This coordination edit creates no implementation claim and no extra receiver.
Current delivery and pairing-card direction — 2026-10-11
- The Human accepted and authorized integration of IM tracer: continuous authorization groups and explicit person exceptions #1406. Product PR #1407 and qualified Provider PR #20 are merged. Design: external users and groups with default Web Human authority #1392/IM tracer: continuous authorization groups and explicit person exceptions #1406 are closed and Project Done; task claims have completion outcomes. Merged-main verification passed. Provider full Windows-suite failures in unchanged image-input/update-service tests remain disclosed; no Provider CI run is inferred.
- Design: reviewer DM pairing cards and one-time Lark/IM codes #1404 records the Human's managed-Role delegation ceiling and confirmed group-only public pairing card, with optional reviewer DM copies when a trusted private route is observed. Native child #1414 implements Role choice → actual authorized actor Approve pairing/Reject → canonical pairing → ask again → fresh model reply. Product Draft PR #1418 and Provider Draft PR #21 are open. Earlier QA does not accept this new behavior; one-time codes remain after first-card feedback.
- Lark tracer: delegated pairing review from group and private-DM cards #1414 remains In Progress in Project M3: Bot 模式 IA 与聊天外壳(列表/Section/DM/本地消息) #10 until merge. Accepted first slice and completed remaining real qualification: actual Reject, delayed group/person Deny, restart, same-card approval after restoration, fresh reply 667, and revoke; old requests did not execute. Exact QA1 receiver stopped/released. Both PRs include sanitized rendered screenshots, baseline capture limits and test results. Native questions, saved rules and management-destination owners/gates remain unchanged; codes remain the next Design: reviewer DM pairing cards and one-time Lark/IM codes #1404 slice.
Earlier dated sections below are history; this current section supersedes their pending-integration and undecided #1404 summaries.
Current delivery — merged 2026-10-11
The Human accepted all four real Lark flows, the complete #1374 first-flow gate, Role/approval and open-group/passive-context interactions, and the latest compact UI with complete PR-body screenshot sequences, then explicitly authorized merge.
| Closed slice | Merged PR / main commit | Exact integrated head / full CI |
|---|---|---|
| #1373 | #1377 / 78a57f1d |
9adbf608 / passed |
| #1374 | #1379 / cf4a9b7f |
213a0314 / passed |
| #1375 | #1387 / 11776eab |
d865ee81 / passed |
| #1376 | #1388 / fe84f5fd |
21cc9d1d / passed |
All four implementation cards are Done. Native dependency edges are retained and fulfilled; the Human-feedback gate was satisfied before #1375/#1376 implementation. Main integration preserved both feature intents and bilingual ledgers; the previously unmerged Lark pairing ADR is 0165, avoiding main's independent ADR-0164. Local integration type/lint/format/build and 89 focused behavior tests passed. Current CodeRabbit statuses indicate rate limiting/skipped stacked reviews, not new independent review evidence.
This Hub stays In Progress/open for the two separate native design follow-ups:
- #1392: external users/groups, default full authenticated Web Human authority. Verified AWS IAM comparison recommends continuous Group grants and separate deterministic capability resolution; permission precedence and the first tracer contract remain unapproved. No Group implementation is claimed.
- #1404: reviewed one-time Lark/IM pairing codes, promoted from the existing Lark tracer: role changes and current-capability approval checks #1375 proposal. Reviewer/delegation authority, command/issuance and delivery contracts remain undecided. No code-flow implementation or prior-QA acceptance is claimed.
Both follow-ups remain needs-triage/Todo, without implementation claims. Project-wide live dependency/ownership audit retains existing IM owners and unresolved real-platform/native qualification. QA1 is stopped and handed back; machine-local reusable App settings remain intact. No release, deployment or receiver restart.
Task: codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76; state: completed for the four reviewed implementation/merge slices. Earlier review-state sections below are historical evidence, superseded by this delivery summary.
Parent
Goal
Deliver the confirmed external-user authorization spec as four production-shaped vertical slices. The Human-approved specification now uses trusted per-message identity and on-demand single-person permission lookup instead of automatic role/permission injection (confirmed revision, 2026-10-11). This hub coordinates delivery, native child relationships, and the first Lark Human acceptance gate; it is not an implementation ticket.
Delivery map
| Slice | Behavior | Blocked by |
|---|---|---|
| #1373 | Lark restricted conversation: ordinary role, automatic pending, sidebar review, re-ask notice, fresh-message reply, on-demand current-permission Tool, revocation | None; reuses delivered #1027 and #1029 |
| #1374 | Bot answers scoped directory queries and completes the spec's first real Lark acceptance flow | #1373 |
| #1375 | Human changes roles and supported approval capabilities; fresh permission lookups and actual native decisions use current authority | #1374, including recorded first-slice Human feedback |
| #1376 | Human opens an exact group to visitors; separately collected unpaired text remains passive context without triggering a model turn | #1374, including recorded first-slice Human feedback |
#1375 and #1376 have no semantic dependency on each other. They may run after the first validated flow, with implementation coordination for shared owner changes. #1374 is a delivery gate for Human feedback, not a claim that a directory query is technically required to render an approval card or open-group toggle.
Tracker placement
- This hub is a native child of Spec: Bot-scoped external user roles and automatic conversation pairing #1371; Lark tracer: ordinary roles and reviewed automatic conversation pairing #1373–Lark tracer: explicit open groups and non-triggering passive context #1376 are its implementation children; Design: external users and groups with default Web Human authority #1392 is a separate native design follow-up. Retain existing task claims and native topology. The Human explicitly authorized the 2026-10-11 spec/ticket revision; it does not authorize implementation takeover or closing the coordination/spec issues.
- Use Project M3: Bot 模式 IA 与聊天外壳(列表/Section/DM/本地消息) #10, IM Identities & Messaging. Implementation claims move cards from Todo to In Progress; review/merge and the recorded Human gates must finish before Done. Implementation tickets use ready-for-agent, area:fullstack, feature:messaging, scope:tracer. Hub uses scope:hub and feature:messaging, and does not represent a coding claim.
- Native blocking links match the table. Keep readable links in each ticket body. No release milestone is inferred from the spec.
- No prefactoring ticket is needed: each slice extends the existing Messaging owner, canonical Operational Database, qualified exclusive Consumer, and authenticated Host/Client service seam. Any small prefactor lands inside the relevant behavior slice and stays green.
Existing work reused
Delivered #1027 and #1029 supply reviewed Lark pairing and checked private approvals. #1019 owns the wider management roadmap; #1031, #1033, #1034, #1046 and their existing dependencies/claims remain unchanged. These new tickets do not implement native questions, save-rules, management group destinations, or other providers.
Confirmed design revision — 2026-10-11
#1373 provides trusted sender/source context and the minimal read-only current-permission Tool; #1374 adds the three directory views. Ordinary chat does not force a query, while stable behavioral guidance asks the Bot to query the actual requester before a new role-governed request. Each lookup reads current state; old Tool Results remain dated evidence. Host admission and supported management checks remain independent of model Tool usage. No new issue, dependency, provider, or hard arbitrary resource ACL is added.
The same task delivered and obtained explicit Human acceptance for all four real Lark slices. Preserve each Issue's original claim and dated QA evidence; current PR heads and outcome comments identify the latest implementation/review state.
UI review and separate user/group follow-up — 2026-10-11
The Human chose to optimize the existing single-Role model first. PR #1377 now uses a wider people/roles/requests table workspace with selected-entry details, contextual information icons and separate Role creation; PR #1387 adds the separate Role editor and single-Role reassignment. Published desktop light/dark and 390 × 844 before/after screenshots are in the PRs. The original real Lark behavior is accepted; the revised layout is available for Human visual review. The four implementation cards stay In Progress while review/merge is outstanding.
#1392 records external users/groups separately in Project #10 Todo with needs-triage. Direct authenticated Web Human input keeps default full Human/Super Admin authority without IM pairing, Role setup or prompts; local-only users need no permission configuration. Messages merely displayed or forwarded in Web do not acquire trusted Web provenance. Group/Role/direct-permission relationships and precedence still require a reviewed design and a runnable first-slice ticket before implementation.
Completion
The four slices together fulfill the spec's first-version behaviors. Every slice includes its real owning Host, durable authority, public service/adapter, appropriate sidebar interaction, focused behavioral coverage, and a runnable demonstration. The first Lark acceptance completes in #1374 before broadening the controls in #1375 and #1376. Mailbox credentials and Himalaya testing remain separate.
Compact UI and complete-flow evidence — 2026-10-11
- PR #1377, head
c1e5dfc9: final-head full CI passed. - PR #1379, head
7de9e58a: final-head full CI passed. - PR #1387, head
5f17c512: final-head full CI passed. - PR #1388, head
5e6ad0e6: final-head full CI passed.
The pairing workspace now aligns portal-scoped native corners, visibly supports search/sort, and shows compact creation/change times from canonical Messaging metadata. Role edits and reassignment preserve creation time; unknown historical fields remain unknown. PR bodies embed the entire entry/creation/save/search/sort and edit/reassignment flows with individually checked before/after, dark and narrow-window screenshots. All four cards remain In Progress pending review/merge; accepted real Lark feedback and native dependencies are unchanged. Both UI-only Hosts are stopped; QA1 was not resumed.
Metadata
Metadata
Assignees
Labels
Type
Projects
- StatusShow more project fieldsTodo
- StatusShow more project fieldsIn Progress
Activity
DoodleBears commentedon Oct 10, 2026
Project #10 dependency, task and Human acceptance audit
Task:
codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76· 2026-10-11 (Asia/Tokyo).Read all Project items, open candidate bodies/comments, native children/blockers and associated PRs. Issue/claim/outcome evidence remains authoritative over a Project card or readiness label. No duplicate issue, owner replacement or Hub implementation claim was made.
Human explicitly arranged QA1 BotHarness Inline Setup QA #1111, stopped its old receiver and granted #1373 an exclusive test window. This is a current window, not an inference from old claim age or a saved idle timestamp. The isolated qualified Provider is connected and the real sidebar has bound QA1, created an ordinary empty-capability Role and enabled restricted reception. The real external request/re-ask/model-reply/revoke demonstration remains pending; no first-flow Human acceptance or downstream gate clearance is claimed.
The reusable QA1–QA3 configuration stays machine-local outside all worktrees. No credentials, raw logs, private Profile or login URL are published. No production takeover, merge, deployment or pin promotion occurred.
DoodleBears commentedon Oct 10, 2026
Current delivery checkpoint — accepted slices and live CI
Task
codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76.Stack order uses the same task's coordinated shared Messaging owner edits; native dependencies remain #1373 → #1374 → {#1375,#1376}. No PR merged/deployed and no card marked Done. Re-read every Project item and refreshed #48/#1019/#1020/#1107/#1144/#1145/#1149/#822 evidence; the previous audit's other owner/blocker/Human gates remain intact. Project Todo and ready-for-agent do not supersede those claims or open capability gaps.
QA1 remains the Human-arranged exclusive receiver window, with one isolated Host/qualified Consumer. Shared test App configuration stays machine-local across worktrees/branches/sessions. #1022's Onboarding QA App is untouched. The proposed one-time IM pairing string remains a separate extension requiring its reviewer/issuance contract; no duplicate issue or authorization by public request references.
DoodleBears commentedon Oct 10, 2026
Review handoff — all four accepted slices
Task
codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76is handed off for review. #1373 / #1374 / #1375 / #1376 each have explicit Human acceptance and passing current-head full CI. PR stack: #1377 → #1379 → #1387 → #1388, all ready for review and registered with T3. #1388 full CI passed 3966 tests (10 skipped), lint/format/typecheck, slides and build; genuine matched before/after group Rules screenshots render in its Change outline.Issue/Project implementation cards remain In Progress until review/merge; no blanket Done or Hub implementation claim. Native issue chain is preserved. Other owners and their open platform/continuation/media/Human QA gates remain as the original audit. No duplicate tickets or automatic merge/deployment.
QA1 BotHarness Inline Setup QA #1111: this task's isolated receiver stopped, process absent and listening port absent; the exclusive window is released for a newly coordinated receiver. Prior receiver was not restarted. Reusable credentials remain machine-local outside all worktrees; the temporary all-text count-2/10s collection policy has been restored to inherited mentions-only, group visitor exception closed, approved ordinary Role retained. No production receiver or #1022 App touched.
DoodleBears commentedon Oct 11, 2026
UI follow-up delivery and current stack evidence
Task:
codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76.The Human chose to improve the existing single-Role UI first. Wide people/roles/requests tables, contextual information icons, selected-entry details and separate Role creation/editing are implemented in #1377/#1387. Desktop light/dark and narrow-window before/after images are published and their GitHub rendering was verified. Real isolated Host creation/edit/reassignment succeeded without a Lark receiver.
All four current-head full CI runs passed; #1377 passed on attempt 2 after one existing 0.1-second CLI timeout harness failure, with no source changes or assertion weakening. The original real Lark interaction acceptance is retained for every slice, including #1374's complete first-flow feedback. Revised UI visual review/merge remains outstanding; all four Project #10 implementation cards stay In Progress, and native issue dependencies remain intact. No automatic merge/deployment or issue closure.
#1392 is a separate native child in Project #10 Todo, needs-triage: default direct authenticated Web Human/Super Admin authority without pairing/permission prompts, and external user/group design with undecided Role/direct-permission precedence. It is not ready for unattended implementation. Existing #1031/#1033/#1034 and other IM owners/gates remain unchanged.
The UI-only Hosts are stopped, task ports released, and QA1 remains stopped. Saved Profiles and reusable machine-local test Apps remain available; another real receiver window must be arranged explicitly.
DoodleBears commentedon Oct 11, 2026
Current delivery — merged 2026-10-11
The Human accepted all four real Lark flows, the complete #1374 first-flow gate, Role/approval and open-group/passive-context interactions, and the latest compact UI with complete PR-body screenshot sequences, then explicitly authorized merge.
78a57f1d9adbf608/ passedcf4a9b7f213a0314/ passed11776eabd865ee81/ passedfe84f5fd21cc9d1d/ passedAll four implementation cards are Done. Native dependency edges are retained and fulfilled; the Human-feedback gate was satisfied before #1375/#1376 implementation. Main integration preserved both feature intents and bilingual ledgers; the previously unmerged Lark pairing ADR is 0165, avoiding main's independent ADR-0164. Local integration type/lint/format/build and 89 focused behavior tests passed. Current CodeRabbit statuses indicate rate limiting/skipped stacked reviews, not new independent review evidence.
This Hub stays In Progress/open for the two separate native design follow-ups:
Both follow-ups remain needs-triage/Todo, without implementation claims. Project-wide live dependency/ownership audit retains existing IM owners and unresolved real-platform/native qualification. QA1 is stopped and handed back; machine-local reusable App settings remain intact. No release, deployment or receiver restart.
Task:
codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76; state: completed for the four reviewed implementation/merge slices. Earlier review-state sections below are historical evidence, superseded by this delivery summary.DoodleBears commentedon Oct 11, 2026
Integration update: #1414 / product #1418 and Provider BotHarness/dsh-im#21 are merged; #1414 is completed. Exact-head product CI and package checks passed. The Human accepted the named Group behavior-permission proposal and selected #1423 as the next task; its own current claim is active. Primary Role and Host capability/delegation boundaries remain separate. Codes #1404, questions #1031, saved rules #1033 and destinations #1034 retain their existing scope and gates. QA1 remains stopped.
DoodleBears commentedon Oct 11, 2026
#1423 first Group behavior-item slice — ready for review
PR #1428 is ready for review; full final-head CI passed. The Human accepted the first real Lark interaction. Current same-ID overlap/Deny resolution, shared-item revisions, affected-person previews and source removal are qualified through the existing Messaging/Operational Database/settings/model-query seams.
The PR body now includes the complete creation-to-final workflow, 17 verified GitHub-rendered before/after screenshots, compact wide Group editing and short-screen body scrolling. Exact-head explanation and evidence.
#1423 remains open/In progress: declared mutual exclusion, incompatible-edit gating and optional semantic advice are later slices. Natural-language items remain prompt behavior; Host management/delegation and Web Human authority are separate. QA1 receiver is stopped and its window released. Task
codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76has handed off the first slice for review; no merge or deployment has occurred.DoodleBears commentedon Oct 11, 2026
Project #10 live audit — 2026-10-11
Source: all Project items plus the live candidate Issues, comments, native sub-issues/blockers and related PRs. Issue/claim/native dependency state takes precedence over Project prose or readiness labels.
73c75f86aa50767be34a5c8d2cadc17b6a3e8862, after accepted Human feedback and final-head CI. Design: reusable group behavior permissions with explicit overlap and conflict visibility #1423 remains open: the delivered same-ID Group behavior permissions/conflict UI is only its first slice. Declared mutual exclusion, incompatible-edit gating and optional semantic advice remain.codex/local/01a12734-8226-7ae3-8d1b-db08bde64d76; native baseline dependencies Lark tracer: ordinary roles and reviewed automatic conversation pairing #1373/IM tracer: continuous authorization groups and explicit person exceptions #1406 are fulfilled. Slack tracer: verified pairing and native Allow once / Reject #1046 remains separate Slack management pairing/native approvals. Slack group parity is not included in Slack tracer: ordinary roles and reviewed chat pairing with current-authority intake #1415.Next gate: finish #1415's focused regression/CI and obtain real Slack pending → Web Role review → fixed re-ask → fresh own-app DM reply → revoke → restart evidence with Human feedback before any group/provider breadth.