Why: cleanup, component synchronization, receipt recording and credential handling could contradict the documented development boundaries.
Evidence: SELF 75 isolated governance tests and source checks are captured in .task/feature-governance-reliability.md; combination and CI verification follow.
Impact: hardens local task and CI tooling, aligns policy and component documentation, and adds continuous regression tests without changing product behavior.
Refs: .task/feature-governance-reliability.md; policy/governance.md; policy/platform-admission.md
Why: Expanding the preview left the conversation column in the row because .main { display: flex } overrode the UA [hidden] rule. Frontend PR #47 restores that hide and centers a wide docx page.
Evidence: Frontend vitest previewfit + design-guard 10 passed. Headless Chromium 1512x900 shows expanded .main display:none, preview slot 1184px, page center delta 0; at 420px and 640px the wrapper stays column-width and the page stays left-aligned.
Impact: modules/frontend 0a0b556 -> 50c7244. Narrow-column fit scaling is unchanged. previewfit still returns 1 when available width is <= 0.
Refs: .task/fix-preview-expanded-hidden.md; https://git-home.u1680832.nyat.app:30319/weichen/Agent-System-Frontend/pulls/47
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Backend PR #62 merged as 3f438b9 (superseded by #64/#65); task card
requires re-pinning the gitlink to the backend merge commit before the
Dev PR can merge.
Evidence: backend go vet + go test ./... -count=1 all ok (29 packages);
merged tree content-identical to backend main.
Impact: modules/backend 685bd0d -> 3f438b9.
Refs: .task/fix-pg-boolean-columns.md
Why: The right-hand preview clipped docx instead of scrolling, and a second overlay viewer duplicated FilePreview. Both fixes are already on the module mains.
Evidence: make test passed in the task worktree (boundaries, pins, libs, core, backend, frontend 82 tests, ui 103 tests, agent-command).
Impact: modules/ui moves to 27d5adb (icons #36 and contrast #37). modules/frontend moves to 0a0b556 (preview scroll #46, including contrast #45).
Refs: Agent-System-UI#36, Agent-System-Frontend#46, Agent-System-Dev#196
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The frontend branch gained two follow-up commits that stop a typecheck artifact from dirtying the submodule.
Evidence: modules/frontend HEAD is 9badbb1 and git status in that repo is clean.
Impact: The Dev branch records the frontend tip that includes page scaling and the tsbuildinfo ignore. No other gitlinks move.
Refs: feature/docx-paper-preview
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The preview painted docx pages with the theme background, so black document text was unreadable in dark mode and the page overflowed the narrow column.
Evidence: UI vitest 84 passed and tsc clean at b14c9c0; frontend vitest 81 passed and tsc -b clean at cf6f24e.
Impact: DESIGN.md documents --paper and --paper-ink. Gitlinks pick up the UI and frontend paper-preview commits. Mammoth and fake-docx text fallbacks stay theme-styled.
Refs: feature/docx-paper-preview
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: the default unarchived task list still compared a BOOLEAN column
to the integer literal 0, so GET /api/conversations returned 500 on
the postgres tier after the previous dialect fix.
Evidence: backend PR #65 (8d82408). Before the fix, TestConversationFlagsAndTaskList
on pg1 failed with SQLSTATE 42883; after, go test ./internal/store -count=1
passed on pg1. Dev make test passed on this pin.
Impact: modules/backend gitlink moves to 8d82408. No other surface changes.
Refs: Agent-System-Backend#65
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: the backend fix unblocks account management, login and conversation
flags on the postgres tier (1/0 literals and int scans rejected with 42804).
Evidence: backend PR #64 merged; live pg1 CLI verified (reset-password +
user list OK); combined make test green.
Impact: dev pins move to the fixed backend; no other surface changes.
Refs: Agent-System-Backend#64
Why: The incremental build info commit has to be the gitlink Dev records.
Evidence: modules/frontend is eb6502f. make build produced bin/agent-system.
Impact: The parent branch names the same frontend commit the worktree built.
Refs: .task/feature-sillytavern-review.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The shared types dropped a spare export so the knip ratchet stays at its baseline, and every container must name those commits.
Evidence: core 0b5c774, ui 6b2589c, frontend 13a1c63, backend c05832d. UI knip 55 and frontend knip 43.
Impact: Gitlinks move together. JSON field names are unchanged.
Refs: .task/feature-sillytavern-review.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dev's agent-native moved for message shaping, so squad and backend have to follow or check-pins fails.
Evidence: Squad 5a57636 and backend include that pin. make test previously stopped at test-pins on this mismatch.
Impact: The parent, backend, and squad all resolve agent-native 3a110a7.
Refs: .task/feature-sillytavern-review.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Empty innerHTML clears were flagged because the space matcher could backtrack past the quotes. The child tests now keep that whitespace inside the lookahead.
Evidence: Frontend vitest 78 passed. UI html-sinks passed. Child commits are ui 5f8308a and frontend 02cebc0.
Impact: The parent pins match the ratchet that allows clearing a node before appending sanitized children.
Refs: .task/feature-sillytavern-review.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The research record, HTML cleaning, listen guard, secret store, and context controls land together, and the parent pointers have to name those commits.
Evidence: Child commits are core bd63223, agent-native 3a110a7, ui d76a8e6, frontend 36b92ed, documents f3f823e, backend 42cca8f. svc.sh forwards the listen switches and info prints them.
Impact: A reverse proxy or a LAN hostname needs allowedHosts or an unauthenticated host check returns 403. Startup moves raw API keys into the secret table after a settings snapshot.
Refs: .task/feature-sillytavern-review.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: advance the backend gitlink to the boolean-column fix so the dev
runtime's PostgreSQL tier stops failing every conversations endpoint.
Evidence: backend branch pushed (685bd0d); make test run recorded in
the task card.
Impact: gitlink-only change in Dev; combined verification via make test.
Refs: .task/fix-pg-boolean-columns.md
Why: the Dev branch must record the Backend main commit that contains the settings fix, not the unmerged feature tip.
Evidence: Backend #61 merged as fd17011.
Impact: modules/backend gitlink moves from d90b9c0 to fd17011. Behavior is the merged fix.
Refs: fix/settings-dotted-keys
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: dotted setting writes were stored as unread keys, and secret pools were returned in clear text.
Evidence: Backend go test ./... at d90b9c0. Dev make test follows this pin.
Impact: the dev branch tracks Backend fix/settings-dotted-keys until that PR merges, then the pin moves to main.
Refs: fix/settings-dotted-keys
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: the shared Python host and the MCP connect change landed in the libraries the product already calls.
Evidence: native, squad, and backend feature commits are recorded on the task card.
Impact: gitlinks move to those commits. Code orchestration stays off unless the setting is turned on.
Refs: feature/code-orchestration
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: tool intent, file checkpoints, fallback, and span names landed in the libraries and the product shell.
Evidence: native, squad, core, backend, frontend, and ui feature commits are recorded in the task card.
Impact: gitlinks move to those commits. Default fallback stays empty and replay stays never.
Refs: feature/durable-rewind
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Session mode, confirmation detail, and the shared tool gate now exist in the libraries the product builds.
Evidence: make test passed. Plan mode rejects write_artifact while the default mode still runs it. Hook adapters match the old stop, reminder, and deny counts.
Impact: Empty permission mode keeps today's profile policy. There is no mode that skips every confirmation.
Refs: feature/permission-hooks
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The conversation screen has to import the phase type so the unused-export ratchet stays at the baseline.
Evidence: frontend 5961b6e, knip 43.
Impact: No other pin moves.
Refs: feature/runstate-shell
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The composer phase and the SSE retry interval have to land together so a reconnect shows 重连中 and still receives the missed events.
Evidence: frontend dcf9b6b, backend 7bef7a4.
Impact: No change to permissions, models, or the already merged live-bubble fix.
Refs: feature/runstate-shell
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Tool-result clearing, folded reminders, and character truncation land in the native loop and the product path together.
Evidence: agent-native 6fd1d44, agent-squad 4d53450, backend b2dc9a7.
Impact: Default context window stays 32000 characters. Empty tool selection still exposes the profile's tools.
Refs: feature/context-engineering
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: New cross-layer imports and unused exports could land unnoticed, and confirm-waiting reused the success color.
Evidence: bash scripts/lint-boundaries.sh passed, including a probe that a new edge fails. Both design-guard tests passed. Knip counts stay at the baselines.
Impact: make test-boundaries now checks architecture edges. Frontend and ui test targets also run the knip ratchet.
Refs: feature/architecture-baseline
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The direct path's default retry only works once every repo builds the same model stream.
Evidence: native 9f5751c, squad 9efce79, backend 93adb93.
Impact: Three gitlinks move. No product UI change.
Refs: feature/llm-robustness
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The stack proposal and the durable note need the v1.0.1 correction before later batches change code.
Evidence: Documents 847f681 adds the second-pass record, the durable correction, and proposal section 2.4.
Impact: Only the documents gitlink moves. No library or product code.
Refs: feature/pi-zcode-second-pass
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The submodule pull requests landed as merge commits, and the parent pin must include those tips so the composed tree keeps both this fix and the commits that arrived on submodule main.
Evidence: Frontend main is 8be98b4 (contains e4ca2ac); Backend main is 28ccb74 (contains 51b1ff6).
Impact: Dev #161 now pins the commits that are already on each submodule main.
Refs: Frontend #38, Backend #54, Dev #161
Why: Sending a message left the composer on 发送 and hid the live process, because the user message now carries the run id and a refresh had no cursor to replay an open run.
Evidence: make test passed in the worktree. A mock server showed the stop and queue buttons within 250ms, the reply growing before it settled, and the in-progress turn restored after switching away and after reload.
Impact: Frontend and backend gitlinks only, plus this task card. No core contract change.
Refs: Frontend fix/live-run-visibility, Backend fix/live-run-visibility
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The review and the four recovery fixes live in the documents, backend, and frontend repositories. The Dev pin is what the combined test runs.
Evidence: Documents 6340900, backend 44aa764, frontend 763c4b4. Each submodule branch is feature/pi-durable-borrow, based on that repository's origin/main.
Impact: Gitlinks only. The pins also move documents, backend, and frontend forward to the main commits that landed after the previous Dev pin.
Refs: feature/pi-durable-borrow
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: the proposal branch has to contain current main before it can land. The owner asked for the open proposal pull requests to be merged.
Evidence: origin/main is 508d216. Documents #34 is merged as 76044b5, same tree as 233dfb54, which this branch already pins.
Impact: the blueprint task card and the documents pin stay. Archived cards follow main.
Refs: Dev #151, Documents #34
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dev has to record the native head that treats a cancelled stream as aborted, and the squad and backend checkouts that point at it.
Evidence: GOWORK=off go test -count=50 -race ./runtime/llm/ passed on d84a67c.
Impact: gitlinks only. The earlier make test covered the previous pin.
Refs: D1
Why: Dev has to test the split modules and refuse a core import of a sibling library.
Evidence: bash scripts/lint-boundaries.sh printed boundaries ok. Full make test follows on this pin.
Impact: test-libs runs the intent, native, sqlite, and command modules. The sibling allowlist is empty.
Refs: D1
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: native 855fe41 is the first head whose go.sum covers intent's NFKC import, so the later pins have to move with it.
Evidence: native go test of runtime, adapters, and exec passed; squad and backend vet passed on 855fe41 and 31ee395.
Impact: three gitlinks and the task card move. Product code in Dev is unchanged.
Refs: I4
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dev must record the intent, native, squad, and backend commits that close this batch so check-pins sees one tree.
Evidence: agent-intent 05e74d8, agent-native 11d1a76, agent-squad 1bb0c3a, and backend a494375 are the feature/agent-stack-r2 heads.
Impact: four gitlinks move and the task card records the batch scope.
Refs: R2 pin
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dev must record the commits that contain the hold fixes, the offline pin check, and the CI checkout change.
Evidence: intent 7f7879b06b053f0a62746309b79e0e1b9a00f42d, native 91c155618487e21edeb4f1d3cc27e985aae2b0d2, squad e928ea3c0fa649cbbbad22a93a1bc0a12dee5f7d, backend ef0a1bef0f7bee626e59f160be51fb0e5f6aa692.
Impact: The four gitlinks match the open pull request heads. Other pins are unchanged.
Refs: feature/agent-stack-r0-r1
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dev could ship a combination whose nested gitlinks differed from the top-level pins, and make test never ran the three agent libraries.
Evidence: A go.work was measured and not adopted: workspace MVS selected newer golang.org/x crypto, exp, net, term, and text than backend with GOWORK=off. check-pins passes on the new pins and fails when backend's squad pin is 5cbac00.
Impact: make test adds pin checks and GOWORK=off vet/test for intent, native, and squad. make build still compiles inside backend; the pin check requires those gitlinks to match Dev.
Refs: R0
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The three agent libraries need one reviewed design for defects, external practice, and decoupling before any further code lands.
Evidence: Documents 233dfb54ea1cd8e8085769832eef18fb3eaa04b7 adds the proposal and the audit record; PR https://git-home.u1680832.nyat.app:30319/weichen/Agent-System-Documents/pulls/34
Impact: Gitlink only. No library code, no runtime change. Later stages stay separate tasks and are not merged here.
Refs: feature/agent-stack-blueprint
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The final reply still printed the artifact id and byte count.
Evidence: go test ./internal/runtime/ -run TestComposeReplyDropsInternalRefs
Impact: Backend gitlink moves to 46d5f7b.
Refs: feature/final-reply-compose
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The first composed reply still included the model's think block and the internal quality score.
Evidence: go test ./internal/runtime/ -run TestCompose
Impact: Backend gitlink moves to 1512ae5. No other module changes.
Refs: feature/final-reply-compose
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The work timeline was being shown as the answer, and file cards appeared before the run had a finished reply.
Evidence: make test
Impact: Gitlinks move to the final-reply branches. The timeline spec now includes the finalizing phase and delays deliverable cards until the reply is done.
Refs: feature/final-reply-compose
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Tool calls stayed invisible while arguments streamed, and a failed run hid the only reply inside the collapsed process block.
Evidence: make test in the worktree passed: boundaries, core, backend, frontend vitest 66, ui vitest 79, agent-command. Mock browser check showed the Chinese process summary, the reply outside the fold, a humanized error with 继续执行, and the back-to-bottom button.
Impact: Dev gitlinks move to the feature commits for core, agent-native, backend, ui, and frontend. DESIGN.md §9.6 describes individual tool blocks for exec, write, and delegate.
Refs: feature/stream-display-overhaul
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: P4 is on main, and the shell pull request has to record UI and frontend commits that contain both the workflow contracts and the shell.
Evidence: submodule trees match the merged mains for core, backend, documents, ui, and frontend.
Impact: the shell branch now contains P4's example workflows and the combined pins.
Refs: feature/s4-shell-overhaul
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The confirm card only folds after the client refetches the decision, so the parent pin has to include that frontend commit.
Evidence: make test passed; mock browser checked the shell, the pending filter, the blocked card, the folded row, and the molv theme.
Impact: task card records the result; modules/frontend moves to the refetch commit.
Refs: feature/s4-shell-overhaul
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The molv theme and the rebuilt shell land together, so the design doc and the submodule pins stay in one change.
Evidence: UI npm test 79 passed; frontend tsc -b and npm test 61 passed.
Impact: DESIGN.md records the molv token table; modules/ui and modules/frontend point at the shell branches.
Refs: feature/s4-shell-overhaul
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dev should record the backend commit that runs a registered script from the launch of that call.
Evidence: go test ./internal/runtime/ -run TestRegisteredWorkflowUsesItsOwnScriptCopy passed.
Impact: Only the backend pointer moves.
Refs: feature/agent-harness-p4
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dev should record the backend commit that makes planning return JSON and makes the critic reject a missing line.
Evidence: go test ./internal/runtime/ ./internal/prompts/ passed on that commit.
Impact: Only the backend pointer moves.
Refs: feature/agent-harness-p4
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The step critic has to call the configured model, and Dev should record that backend commit.
Evidence: go test ./internal/runtime/ -run TestJudgeUsesConfiguredModel passed in the backend repo.
Impact: Only the backend pointer moves.
Refs: feature/agent-harness-p4
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dev has to record the Core, Backend, Frontend, UI, and Documents commits that implement intent-to-workflow.
Evidence: make test passed, including boundary lint, core, backend, frontend vitest, UI tsc and vitest, and agent-command.
Impact: agent-native stays at a34689c. The example plugin is already on this branch.
Refs: feature/agent-harness-p4
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The combination has to move together so background subagents, review, and the notice row are the same build.
Evidence: Each module's own tests passed on its feature branch. agent-native stays at a34689c.
Impact: Dev does not merge this pin. The four module pull requests stay open until review.
Refs: feature/agent-harness-p3b
Co-authored-by: Cursor <cursoragent@cursor.com>
weichen
