Why: src/contracts.ts is regenerated from Agent-System-Core by the Dev
composite; the scratch field added there is dropped on regeneration and
tsc then fails in make build.
Evidence: contracts.ts equals the Core-generated output; npx tsc -b
--noEmit, npx vitest run (685 tests) and check-knip pass locally (SELF).
Impact: ConversationView (Conversation & { scratch?: boolean }) in
api.ts carries the server-local flag, mirroring the Backend
conversationView; no behavior change.
Refs: Agent-System-Frontend#123, Agent-System-Dev#691;
agent-exec receipt u3f-run (cursor, success)
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Independent review found the cross-read toggle, the project inbox
scope and the archive-to-project action still reachable from an open
scratch (no-project) conversation.
Evidence: npx vitest run (685 tests) and npx tsc -b --noEmit pass locally
against the UI fix branch (SELF); CI on the PR head is authoritative.
Impact: Scratch conversations hide the cross-read setting, always show
the conversation inbox and omit onArchive; real projects are unchanged.
Refs: Agent-System-UI#60; agent-exec receipt u3e-run (cursor, success)
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Accounts without project.create need isolated general-agent chats.
The sidebar groups every scratch session under one "无项目对话" entry
instead of the project switcher, and the client no longer auto-creates
a default project.
Evidence: npx vitest run (64 files, 677 tests) and npx tsc -b --noEmit
pass locally (SELF); CI on the PR head is authoritative.
Impact: createTask uses POST /api/scratch-sessions in scratch mode or
when no real project exists; scratch rows hide cross-read toggles and
project settings; "新建项目…" stays gated by project.create.
Refs: Agent-System-Dev docs/superpowers/specs/2026-10-22-scratch-session-general-agent-design.md;
agent-exec receipt u3-run5 (cursor, success)
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Synchronize Frontend main (IAM U3 account UI #86) into the super-admin branch and resolve conflict in api.ts, AccountUsersPanel.tsx and adminSections.tsx as unions.
Evidence: npm test passed (49 files, 574 tests), check-knip passed.
Impact: AccountUsersPanel supports both super-admin identity/elevation and U3 initialization/credential states.
Refs: #627, Frontend PR #86
Why: U7 requires structured regulation citations to be visible in audit and document views.
Evidence: CI success on 7050c91334 (governance); vitest/knip/tsc/vite build passed in the composed layout (SELF).
Impact: New regs audit tab and citation rendering; existing tabs unchanged.
Refs: https://git-home.u1680832.nyat.app:30319/weichen/Agent-System-Frontend/pulls/109
Why: at 390px the single-pane layout lands on 对话 after login and hides
the whole context sidebar (both sidebar and rail forms); scenario 7 must
switch to the 上下文 pane via the top bar before the entry exists.
Evidence: tsc clean, vitest green, build green (SELF); seven-scenario
rerun to be recorded on the Dev I4 card.
Impact: e2e spec only.
Refs: rollout I4
Why: at 390px the compact layout swaps the sidebar for an icon rail; the
cockpit entry remains reachable as an icon button but not as .tsb-tool,
so scenario 7 timed out waiting for the sidebar form.
Evidence: tsc clean, vitest green, build green (SELF); seven-scenario
rerun to be recorded on the Dev I4 card.
Impact: e2e spec only.
Refs: rollout I4
Why: the cockpit refresh reloaded overview and tree but left the node
detail at its old snapshot, so the 刷新对齐 banner persisted after
refreshing (DV-R05 wants every projection to advance). PanoramaNode now
refetches when the overview snapshotRevision advances. Scenario 4 also
scoped its tree clicks to the 验证主体报告 branch — the unscoped first
caret expanded an unrelated rdm-self doc.
Evidence: tsc clean, vitest green, build green (SELF); seven-scenario
rerun to be recorded on the Dev I4 card.
Impact: cockpit panorama + e2e spec.
Refs: rollout I4
Why: scenario 4 relied on a pending draft proposal advancing the
snapshot watermark, but rdm.draft.propose commits nothing (wm:58 before
and after) and draft adoption is high-danger (client cannot forge
Confirmed). rdm.reqs.import with a unique probe item is a committed
low-danger mutation that advances the watermark — the honest DV-R05
driver.
Evidence: tsc clean, vitest green, build green (SELF); wm 58->60 probe
against the seeded data recorded in session.
Impact: e2e spec only.
Refs: rollout I4
Why: the I4 scenarios assert the version collapsible for a selected
section in the panorama (not only the reader); the panorama node view
had no version block, so .cockpit-versions-toggle never existed there.
Also normalize null list projections in the panorama node detail fetch
(same Go nil-slice issue fixed for the panel earlier).
Evidence: tsc clean, vitest green, build green (SELF); seven-scenario
rerun to be recorded on the Dev I4 card.
Impact: cockpit panorama only.
Refs: rollout I4
Why: rdm.cockpit.node projects empty conflicts as items:null (Go nil
slice) while sources arrive as []; the panel read items.length on null
and crashed, so the conflicts empty state never rendered (I4 e2e
scenario 3).
Evidence: tsc clean, vitest green, build green (SELF); seven-scenario
rerun to be recorded on the Dev I4 card.
Impact: cockpit panel only.
Refs: rollout I4
Why: the local ProjectView interface invented 'id'; rdm.project.list
returns projectId. Every option value was undefined, so auto-select set
pid to undefined and the overview request was never issued — cockpit
showed 加载概览… forever after the project list loaded (I4 e2e scenario 1).
Evidence: tsc clean, vitest green, build green (SELF); network probe shows
rdm.cockpit.overview 200 and .cockpit-stats rendered; seven-scenario rerun
to be recorded on the Dev I4 card.
Impact: cockpit entry only.
Refs: rollout I4
Why: the cockpit rendered the project dropdown with no project selected
and an empty main area — the design's 首屏 default (first visible project
drives the overview) was never implemented; caught by I4 e2e scenario 1
which got past the project list but never saw .cockpit-stats.
Evidence: tsc clean, vitest green, build green (SELF); seven-scenario
rerun to be recorded on the Dev I4 card.
Impact: cockpit entry only.
Refs: rollout I4
Why: #102 rewrote the bare-array fallback into a multi-line callback and
turned 'if (!c) setProjects(...)' into an early 'if (!c) return', which
skips setProjects on the normal (non-cancelled) path — the cockpit then
hangs on 加载中 forever, caught by I4 e2e scenario 1.
Evidence: tsc clean, vitest 324/324, build green (SELF); seven-scenario
browser suite rerun recorded on the Dev I4 card.
Impact: cockpit entry only.
Refs: rollout I4
Why: the forced password gate still submitted an empty current password
after the backend began requiring it (current_required), locking every
account seeded with an initial password out of the workbench, and the
admin user panel had no client for the owner-confirmed recovery
proposal contract that replaced one-time key issuing (HTTP 410).
Evidence: npm test 300 passed; tsc -b && vite build ok; knip issues
54 <= baseline 55 (SELF).
Impact: mode-aware ChangePasswordGate with current-password input,
admin recovery proposal panel, owner-side account settings section,
recovery proposal client methods; no backend changes.
Refs: backend 8dfc223, b9c0a34; frontend #94
Why: independent review flagged the issue client still targeting the 410
endpoint and a hint sentence ending with a period.
Evidence: npm run build + npm test re-run on the fix (SELF).
Impact: api client surface only.
Refs: review brief /tmp/review-brief-admin-user-contract.md
Why: backend 8dfc223 replaced POST /api/admin/users with strict
{owner, profile} decoding and returned 410 for login-key issuing, so the
UI create form failed with 400 unknown field name. Creation now sends
{owner, profile:{name}} only; password input, admin checkbox and the
login-key issue entry are removed; notice points to the local CLI
initialize-account flow. passwords.ts lost its only consumer.
Evidence: npm run build + npm test (34 files/293 passed) + knip identical
to baseline on the task worktree (SELF).
Impact: UsersPanel and api client only; no style/token changes.
Refs: backend 8dfc223, b9c0a34
Why: S9 acceptance finding 3 — the project settings modal exposed the
host override settings (including the model/key group, masked) to plain
project members; the duties data plane was already permission-enforced,
but the host settings surface had no gate.
Evidence: vitest 34 files / 293 tests pass with the change; tsc reports
no errors in the modified file. Ruling per the acceptance plan P1-B:
the modal entry stays (granted duty managers need the RDM tabs), while
the host override settings group renders only for the project owner or
a system admin; RDM tabs keep their own fail-closed gating.
Impact: ProjectSettingsModal project tab renders an explanatory empty
state for non-privileged viewers; members/duties/matrix tabs unchanged.
Refs: S9 acceptance report P1-B
Why: Approving an extra package, rebuilding the base layer, reclaiming unused layers, and uploading a wheel need an admin surface that uses the existing admin controls.
Evidence: SELF tsc of the new panel reported no errors in frontend sources. UI package type errors come from missing node_modules in this worktree, not from this panel.
Impact: Administrators get a Python 环境 section. Other admin sections are unchanged.
Refs: .task/feature-layered-python-env.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: a lost write response was keeping the old duty version, so the next save replayed it, and a busy save could stay locked if another request advanced the sequence.
Evidence: SELF vitest duty-draft 6 passed, tsc passed, knip 43, Playwright duty flows 3 passed.
Impact: an unknown save reloads server roles, confirmation resets, and busy clears even if a later read is discarded.
Refs: docs/proposals/2026-10-15-rdm-duty-matrix-rework.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: project settings must keep the shared duty draft and must not regain template document creation from main.
Evidence: SELF vitest 292 passed, tsc --noEmit passed, knip unused exports 43.
Impact: the duties tab stays assignment-only while conversation ordering from main is included.
Refs: docs/proposals/2026-10-15-rdm-duty-matrix-rework.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Wait for the unrelated foreground reply before measuring duplicate replies.
Evidence: All 24 scenarios passed against deployed 8788; independent replay passed the targeted scenario.
Impact: Test synchronization only; the duplicate-reply assertion is unchanged.
Refs: Agent-System-Frontend #89; Agent-System-Dev #565
Why: Completion or progress from an earlier task could appear in the process of an unrelated newer answer.
Evidence: The new ownership regression fails before the guard and passes afterward; the frontend suite, TypeScript and unchanged export gate pass.
Impact: Project child progress only into its originating live turn while still reconciling background results.
Refs: feature/workbench-flow-completion chronological display requirement
Why: Full execution history must retain outputs after the 500-event API page boundary.
Evidence: Ordered 502-event replay and nonadvancing-cursor rejection tests pass.
Impact: Add regression coverage only; production source is unchanged.
Refs: feature/workbench-flow-completion review testing gap
Why: the settings page only needs the draft operations the views call. Extra aliases and unused helpers pushed the unused-export count over the baseline.
Evidence: SELF node scripts/check-knip.mjs reported 43 <= 43, and duty-draft vitest passed.
Impact: chapter-review command aliases are gone from the settings gate. The draft behavior is unchanged.
Refs: modules/frontend/knip-baseline.json
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: A rejected result-reference request could discard the user text or uploads after the composer cleared.
Evidence: 286 frontend tests, TypeScript and the unchanged unused-export gate pass; mixed-input browser regression preserves the draft.
Impact: Validate incompatible inputs before clearing, retain failed pending input with its request identity, and require explicit retry while other messages can continue.
Refs: feature/workbench-flow-completion review frontend-races finding
Why: the three views, the blocked empty-reviewer save, and the narrow-screen matrix need a browser check that does not depend on a live account.
Evidence: SELF playwright e2e/rdm-duties.playwright.ts, 3 passed against the local Vite server.
Impact: the duties end-to-end file now drives assignment, stale responses, and the narrow-screen guidance.
Refs: feature/rdm-duty-matrix
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: the settings page was mixing duty assignment with whole-document review and chapter objects. The three views need one draft and one save check.
Evidence: SELF vitest run of the frontend suite, 278 tests passed, including duty-draft and design-guard.
Impact: the duties tab assigns existing documents only. Submit, approve, reject, and final stay out of this page. Commands that are not ready keep the tab hidden.
Refs: feature/rdm-duty-matrix
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The frontend unused-export gate rejected a helper used only within its module.
Evidence: The unchanged knip baseline and TypeScript checks pass.
Impact: Remove an unnecessary export without changing replay behavior.
Refs: feature/workbench-flow-completion
Why: Keep persisted task decisions aligned across Go and TypeScript.
Evidence: make contracts generated both consumers from Core 1a5ef6b.
Impact: Include the reserved run identity and typed dispatching state.
Refs: feature/workbench-flow-completion
Why: Repeated progress and coupled scrolling obscured chat, and task choices lacked a reliable independent interaction path.
Evidence: 285 frontend unit tests and 23 browser scenarios pass locally before the shared event-loader extraction; final combination checks follow in Dev.
Impact: Coalesce cumulative progress, retain execution records, show timestamps and quiet notifications, support versioned work actions and explicit result references without automatic chat backfill.
Refs: feature/workbench-flow-completion
Why: Retired edges and compliance edges awaiting review both fell back to a generic lifecycle label.
Evidence: Presentation tests, production build, knip, and diff checks pass.
Impact: The trace network now shows 已停用 and 待复核 as distinct human-readable states.
Refs: readable trace network
Why: Consume the Core work result and optional decision snapshot.
Evidence: Generated with make contracts from Core 00f1826.
Impact: Add optional types without changing existing cards.
Refs: feature/workbench-flow-completion
Why: Work-source decomposition, detected risks, human answers, and structure decisions were recorded but absent from the audit timeline.
Evidence: Focused presentation/tree tests, production build, and knip baseline pass.
Impact: Audit filters and rows now include the complete business-operation feed without raw payloads.
Refs: TT-N03 TT-R01 W4b
Why: Object details render the same relationship row and must identify the endpoint relative to that object.
Evidence: Type checking identified the missing selected-object context after the endpoint fix.
Impact: Relationship rendering is consistent in the trace view and object details.
Refs: rdm-two-trees-latest-templates
Why: The trace view selected an endpoint by lexical ID order and could present the current section as its own source.
Evidence: Focused presentation tests cover both selected directions of a source-to-section relationship.
Impact: Source and impact columns now identify the object opposite the selected node.
Refs: rdm-two-trees-latest-templates
Why: Recursive template nodes without titles were rendered as repeated generic project objects in the audit tree.
Evidence: Focused presentation tests pass, including a regression for stable readable fallback names.
Impact: Template-derived paragraphs, tables, rows, cells, fields, and attachments remain distinguishable in the document tree.
Refs: rdm-two-trees-latest-templates
Why: Project document and knowledge structures can contain nested nodes whose children must remain navigable beyond the first section level.
Evidence: A RED test first showed child-bearing sections were not expandable; presentation and design-guard tests, knip baseline 43, and npm run build now pass.
Impact: Recursively render any node marked with children while preserving current document fallback and readable node states.
Refs: TT-N01, TT-N02, TT-D01, TT-U01; two trees one graph vertical slice
Why: Project members need to understand who changed what, why, and with what impact without reading operation codes or raw audit payloads.
Evidence: A RED presentation test first failed on the missing humanization layer; 276 frontend tests passed, design guard 10 checks passed, knip stayed at baseline 43, and npm run build passed.
Impact: Humanize tree, relation, state, and operation vocabulary; render structured reason, impact, source, and outcome fields; raw event detail is no longer shown in the audit timeline.
Refs: TT-U01, TT-U02, TT-U03, TT-U04; two trees one graph stage-five plan
Why: Independent review found component-local color derivation, a private transition, and missing dark/molv coverage in the visual candidate.
Evidence: Frontend design guard passes and 19 Playwright scenarios cover light, dark, molv, and 320-1920px layouts.
Impact: Keeps the approved visual direction while moving color ownership to @ui tokens and removing private motion.
Refs: Agent-System-Dev feature/workbench-visual-redesign
Why: The inherited light-theme divider and hover treatment disappeared inside the new inverse create button.
Evidence: Focused browser layout test and production build pass locally.
Impact: Changes only the create action divider and hover surface.
Refs: Agent-System-Dev feature/workbench-visual-redesign
Why: Spacing-only refinements still looked like the original flat gray workspace and did not satisfy the approved visual redesign.
Evidence: 272 unit tests, 16 browser scenarios across 320-1920px, and the production build pass locally.
Impact: Gives navigation, authored messages, assistant results, and work details clearly different visual surfaces while preserving behavior.
Refs: Agent-System-Dev feature/workbench-visual-redesign
Why: The three-column workspace read as flat and visually scattered, making conversation, current work, and navigation difficult to distinguish.
Evidence: 272 unit tests, 16 focused browser scenarios across 320-1920px, and the production build pass locally.
Impact: Refines the existing workbench presentation without changing its data flow or fast/slow task behavior.
Refs: Agent-System-Dev feature/workbench-visual-redesign
Why: Status words shared with a final reply could suppress a child completion record.
Evidence: 272 frontend unit tests, build, and 15 isolated browser scenarios.
Impact: Tag client-side collaboration replay so partial-delta deduplication preserves agent notes.
Refs: Agent-System-Dev feature/work-companion-visibility UI1
Why: A dropped delta batch can leave a short event prefix next to the complete saved message.
Evidence: 271 frontend unit tests, build, and 14 isolated browser scenarios.
Impact: Suppress overlapping partial replay text and thinking while retaining saved content.
Refs: Agent-System-Dev feature/work-companion-visibility UI1
Why: Event replay could hide saved retry notes and leave completed cards outside the active work filter.
Evidence: 270 frontend unit tests, build, and 14 isolated browser scenarios.
Impact: Merge replay into saved history, retain child progress, and locate cards under the right scope.
Refs: Agent-System-Dev feature/work-companion-visibility UI1 UI2
Why: Fast and slow channel cards duplicate across panes and completed replies hide process output.
Evidence: npm test; npm run build; 12 isolated Playwright scenarios with reload and responsive widths.
Impact: Preserve ordered folded process history, load durable run events, and give work cards one detailed pane.
Refs: Agent-System-Dev feature/work-companion-visibility UI1 UI2 UI3
Why: Users must see truthful first replies when work queues or cannot start.
Evidence: RED and GREEN HTTP browser receipts plus Frontend 269 tests and production build.
Impact: Adds two task-owned Playwright scenarios; production UI is unchanged.
Refs: U3 r0.5 independent review finding
weichen
