Skip to content

test(gui-user): fix nine catalog entries that name UI absent from main - #19683

Merged
iamwhatever merged 1 commit into
mainfrom
fix/gui-catalog-stale-entries
Oct 13, 2026
Merged

iamwhatever merged 1 commit into
mainfrom
fix/gui-catalog-stale-entries

Conversation

@javenciu

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: Make the GUI user-test catalog describe the dashboard that is on main, so a lane that follows the nine table-A entries of #19582 is not sent to routes, controls or titles that do not exist.

An agentic run over the catalog (#18614) left 86 entries red. #19582 triaged 35 of them as catalog or seed errors, and nine of those (its table A) name UI that is not on main or that lives elsewhere, so the lane can only fail them. Each of the nine was re-checked against origin/main at de2d143 with git grep; the receipts are under What changed.

Why it matters

The catalog is the backlog the GUI user-test lane works from. A record that sends the lane to a control the dashboard does not draw can only fail, and the failure reads as a product bug next to the real ones in the nightly report.

Not a goal

  • Parts B and C of GUI catalog: 35 features.json entries fail on wrong routes or missing seed data, not product bugs #19582 (records whose seed fixture lacks data, an enabled app or a config flag). They stay open on the issue.
  • Deleting the unmounted McpInfoButton component. The issue offers that as an alternative; this PR only stops the catalog from pointing at it.
  • Other records that still say "Crew Members" for the rail or the Feature Previews toggle, and the members registry title.
  • Scenarios, the validator or the renderer.

What changed (motivation → approach → change)

Motivation: several surfaces moved after the records were written, and nothing re-reads a record against the code. Approach: for each record, git grep the route or control it names on origin/main. Where it lives under another route or name, correct entry_path, title, notes and source (ids and merged_from stay). Where nothing exists, mark the record excluded with a reason, the catalog's own convention for flows the lane cannot drive (excluded is P3, so the record moves to the Not generated table). One line per record:

  • chat-mcp-tools-panel: McpInfoButton is mounted nowhere, it is only re-exported at website/src/pages/chat/index.ts:5. McpToolsPanel renders at website/src/pages/chat/ChatPageMessageContent.tsx:155, inside the MCP servers submenu of the chat header's session options menu. The path and notes say so.
  • sidebar-crew-members-create-menu-entry: the entry exists but reads "Crewmates" (open_crew_members at website/src/i18n/locales/en.manual.json:5667). The menu item is data-testid="open-crew-members" at website/src/pages/ChatSidebar.tsx:4729 and openCrewMembers at website/src/pages/chat-sidebar/create.ts:191 picks /members or the Feature Previews card.
  • chat-knowledge-picker-attach: there is no picker button. KnowledgePicker mounts at website/src/pages/ChatPage.tsx:6409 only while a prefix search runs, started by the composer intercept (extractKnowledgeQuery at website/src/pages/ChatPage.tsx:2081) for @knowledge , @kb and /kb (PREFIXES at website/src/pages/chat/useKnowledgeFetch.ts:27). The notes add that the memory-populated seed holds memory files and no knowledge source, so the lane must add one first.
  • chat-rewind: git grep -i rewind origin/main -- website/src/pages/chat/AssistantMessage.tsx is empty. Rewind is handleEditResend at website/src/pages/ChatPage.tsx:3347, reached from Edit & Resend on a user message (a DropdownMenuItem at website/src/pages/chat/UserMessage.tsx:697, or the startEdit pencil button at website/src/pages/chat/UserMessage.tsx:730) and served by api_chat_slot_rewind at src/kiro_crew/dashboard/routes/sessions.py:76. The Rewind row of docs/feature-map/README.md named the same missing message action and is corrected too.
  • developer-mcp-pool: the route is right, the tab is MCP Management. The mcp-pool key at website/src/pages/DeveloperPage.tsx:45 renders McpManagement with a Servers table (stub switches) and a Sharing assessment view; it has no pool-state or probe control. Title, story, path and rationale are rewritten around those.
  • kas-auth-sign-in-card-signed-out: website/src/pages/settings/KiroSignInCard.tsx does not exist. The card is website/src/pages/developer/KiroSignInCard.tsx, mounted only in the KAS row of Agent Harness (KiroSignInCard at website/src/pages/developer/AgentBackendTab.tsx:1550), linked as KIRO_SIGN_IN_PATH at website/src/pages/developer/kiroSignInLink.ts:25; docs/system-specs/modules/kas-auth.md says it is deliberately not on Overview. start_url becomes /settings/agent (the validator rejects a query there). settings-overview and memory-browser repeated the Overview claim and get the same correction; Overview mounts only the pointer at website/src/pages/OverviewPage.tsx:334.
  • Two records outside table A carried the same stale card location in their notes and source list, memory-browser and settings-overview; their text is corrected to match, nothing else about them changes.
  • standalone-logs: /logs is a Route in the shell's route list at website/src/App.tsx:2839, and the only popout routes are chat, artifact and terminal (website/src/App.tsx:1736). The story no longer promises a chrome-less page.
  • sidebar-folders-history-and-archive: the fixture session research-notes is titled "Research: MCP transport trade-offs" (src/kiro_crew/tests_fixtures/rich/sessions/dashboard_research-notes.jsonl:1) and the bottom pane is labelled Older Sessions (older_sessions_2 at website/src/pages/ChatSidebar.tsx:6968), not History. The shipped scenario sidebar-folders-and-older-sessions already uses both.
  • sidebar-workspace-switcher-three-workspaces: git grep -n -i -E 'WorkspaceSwitcher|switch_workspace|switchWorkspace|workspace switcher' origin/main -- website/src src/kiro_crew ':!website/src/i18n/locales' finds nothing. Named workspaces are edited in Developer > Config and picked per crewmate, so the record becomes excluded/P3 with that reason.

FEATURES.md is regenerated with python test/gui_user/features_catalog.py --write. The counts block moves nightly 167 → 166 and excluded 22 → 23, and the informational priorities and per_feature blocks follow.

Backwards compatibility

Compatible: every record keeps its id and the validator and renderer are unchanged. One record changes tier, so it leaves the runnable tables.

Reader: test/gui_user/features_catalog.py:render -- release -- FEATURES.md is regenerated in this commit and test_features_md_is_rendered_from_features_json holds the pair together
Reader: test/gui_user/test_features_catalog.py:test_every_p0_is_runnable_on_the_target -- release -- the record is P3, which that test requires of every excluded record

git grep -F for each of the nine ids on origin/main finds only features.json and FEATURES.md, except developer-mcp-pool, which website/scripts/lib/i18n-surfaces.mjs also uses as an unrelated surface key.

Tests

In short: test/gui_user/test_features_catalog.py passes 3 of 3 on the head (35 tests), fails when FEATURES.md is not regenerated from the edited JSON, and all of test/gui_user (305 tests), the 16 ratchet files (234 tests), docs-lint, the feature-map gate and the brand gate are green.

  • Data-only change, so no new test code. test_features_catalog.py already holds the schema, registry slugs, record order, counts, "every excluded record is P3", shipped seeds, no cross-slug duplicate (the new component values on chat-rewind and developer-mcp-pool pass it) and FEATURES.md == render(load()).
  • Negative control: the edited JSON next to the old FEATURES.md gives 1 failed, 34 passed (test_features_md_is_rendered_from_features_json); the regenerated file gives 35 passed.

Manual verification

N/A for a UI run: no source or UI file changes. Every claim above was read back from origin/main with git grep or git show. For two records the shipped scenarios (auth-sign-in-card-signed-out, sidebar-folders-and-older-sessions) were already written against the corrected paths.

Related Issues

Refs #19582

Adjacent open work: #19648 adds a catalog record and edits the same counts block of features.json and FEATURES.md; #19537 edits scenarios. Neither record is touched here. Whichever merges second regenerates FEATURES.md.

Pattern harvest

Rule candidate: lint
Pattern: a catalog record whose source or component path no longer exists in the tree. Before this change 3 of the 270 records cited a missing file; this PR fixes two of them (the sign-in card moved) and capabilities-crews-templates-skills-tabs (website/src/pages/AgentsPage.tsx) is the third. A check in test_features_catalog.py that every such path exists would have caught the move.

Checklist

  • At most two commits (one), with a Conventional Commits title
  • Existing tests pass; data-only change, so no new tests
  • Self-review completed; code follows project style guidelines
  • Documentation updated (FEATURES.md regenerated, feature-map Rewind row corrected)
  • No secrets, credentials, or internal references in the diff

Nine records in features.json (table A of #19582) sent the GUI user-test
lane to routes, controls or titles that are not on main. Each now names
what the dashboard draws, and FEATURES.md is regenerated from the JSON.

- chat-mcp-tools-panel: the panel is in the session options menu, there
  is no chat footer button
- sidebar-crew-members-create-menu-entry: the entry is labelled Crewmates
- chat-knowledge-picker-attach: the picker opens from an @kb, @knowledge
  or /kb search typed in the composer
- chat-rewind: reached through Edit & Resend on a user message
- developer-mcp-pool: the tab is MCP Management (tab key mcp-pool)
- kas-auth-sign-in-card-signed-out: the card is on Settings > Agent Harness
- standalone-logs: /logs renders inside the shell, with its chrome
- sidebar-folders-history-and-archive: the pane is Older Sessions and the
  closed session has its real title
- sidebar-workspace-switcher-three-workspaces: no switcher exists on main,
  so the record is excluded

settings-overview and memory-browser repeated the sign-in card claim and
get the same correction. The Rewind row of docs/feature-map/README.md
named the same missing message action.

Refs #19582
@javenciu
javenciu requested review from a team and KiroCrewCaptain as code owners October 13, 2026 05:11
@javenciu
javenciu requested a review from mbriones98 October 13, 2026 05:11
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 13, 2026
@github-actions

github-actions Bot commented Oct 13, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ⏭️ skipped

Revision 14730ef57b832dbb810a2303c27d8b1c03451886 ships no reviewable capability (docs, tests or generated files only), so there is nothing to inventory. Advisory — does not block merge.

@github-actions

github-actions Bot commented Oct 13, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 14730ef57b832dbb810a2303c27d8b1c03451886 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Nothing to check.

[DESIGN-REVIEWED] 14730ef

Description read

[DESCRIPTION-READ] 6fb64935610a79db55ed1ba8efb05287cee6887564c9ac38442afabe166ea812

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 13, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ⏭️ skipped

Revision 14730ef57b832dbb810a2303c27d8b1c03451886 touches no user-facing surface (no changes under website/ or committed screenshots), so the UX review was skipped. Advisory — does not block merge.

@github-actions

github-actions Bot commented Oct 13, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 14730ef57b832dbb810a2303c27d8b1c03451886 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 14730ef

Verdict parsed from the review's SHA-scoped output markers for commit 14730ef57b832dbb810a2303c27d8b1c03451886.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 14730ef57b832dbb810a2303c27d8b1c03451886: <one-sentence reason>

@github-actions

github-actions Bot commented Oct 13, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 14730ef57b832dbb810a2303c27d8b1c03451886 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 14730ef

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 14730ef57b832dbb810a2303c27d8b1c03451886: <one-sentence reason>

@javenciu

Copy link
Copy Markdown
Contributor Author

CI on this PR has been queued for six hours across three attempts. The same-repo routing sends my runs to the fleet (labels codebuild-kirocrew-gha-linux-<run>-<attempt>, attempts 1 to 3), and no fleet runner has picked one up, while other same-repo runs started on the fleet within a minute this morning. The hosted lanes on the same head completed normally.

The changes job's own comment in ci.yml names this shape: an actor admitted by CODEBUILD_ACTOR_IDS whose jobs the fleet webhook does not accept. A re-run keeps the original actor and event, so it stays on the fleet with a new attempt label; a reopened event is not opened/synchronize and takes the hosted path, which is the documented fallback. Closing and reopening for that reason; nothing else changes.

Maintainers: could the fleet webhook admit my actor, or drop me from CODEBUILD_ACTOR_IDS so same-repo heads fall to hosted on the first try? Either works for me.

@javenciu javenciu closed this Oct 13, 2026
@javenciu javenciu reopened this Oct 13, 2026
@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 13, 2026
@javenciu
javenciu enabled auto-merge (squash) October 13, 2026 12:36

@iamwhatever iamwhatever left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This PR corrects the nine table-A records of #19582 in the GUI user-test catalog (features.json, the rendered FEATURES.md and one feature-map row) so the lane is no longer sent to routes and controls that are not on main; that is the root cause of those red entries, since the records were stale, not the product. I checked the claims against current main: McpInfoButton is only re-exported from pages/chat/index.ts and mounted nowhere, KiroSignInCard lives under pages/developer (no pages/settings copy exists), the menu label is "Crewmates" in en.manual.json, AssistantMessage.tsx has no rewind action, /logs is a plain Route in App.tsx, and no workspace switcher component exists in website/src. Moving the workspace-switcher record to excluded/P3 follows the catalog's own convention, and the counts in both the JSON and the Markdown tables move consistently (nightly 167->166, excluded 22->23, P2 162->161, P3 61->62). The change is data and docs only, so nothing at runtime changes. The red Backend/Frontend/Lint/Build entries are the cancelled fleet run 38350163424; the hosted re-run 38390769127 on the same head is green across every shard, and all five AI lanes pass or skipped with no findings. — crew-pr-reviewer (deep re-check)

This was referenced Oct 14, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants