Skip to content

feat(skills): surface waiting generated skills on the Skills rail row - #18972

Merged
cruisercohen merged 1 commit into
mainfrom
feat/5643-surface-pending-skills
Oct 11, 2026
Merged

cruisercohen merged 1 commit into
mainfrom
feat/5643-surface-pending-skills

Conversation

@cruisercohen

@cruisercohen cruisercohen commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: A user can see that generated skills are waiting for their approval without first opening the Skills tab, and can tell a brand-new skill from an update to an existing one.

Kiro Crew can draft skills from finished sessions and stage them in a review queue for approval. That queue renders only inside Agent Capabilities > Skills, and the page opens on Crewmates (or the last tab used). Nothing anywhere else on the page says a candidate is waiting, so generated skills sit unseen until the user happens to open the Skills tab. Inside the queue, an update proposal carries an "Update" badge, but a candidate that would add a new skill carries no marker at all.

Why it matters

A generated skill does nothing until a person approves it. If the user never notices the queue, the feature silently produces nothing, and candidates age out after the pending TTL without ever being seen.

Not a goal

  • Changing how candidates are generated, deduplicated, validated or approved. The review queue stays the one adoption path.
  • A toast, banner or notification-center entry for new candidates.
  • A badge on the main navigation rail outside the Agent Capabilities page.
  • Any backend or API change.

What changed (motivation → approach → change)

Gap: the waiting count exists (the queue fetches it) but is only drawn inside the Skills tab. Approach: put the count where the user already looks for capabilities, on the Skills row of the Agent Capabilities rail, and mark new candidates in the queue.

Options considered:

  • Chosen: a count pill on the Skills rail row, plus a "New" badge on new candidates beside the existing "Update" badge. It reuses the existing queue as the only place to act, adds no new surface, and reads the same cache entry the queue reads.
  • Rejected: list generated skills directly in the installed-skills list with an inline enable toggle. That bypasses the review step that keeps unreviewed (and possibly script-bearing) skills from going live, and duplicates the approve flow.
  • Rejected: a notification or toast for each new candidate. It is noisier, fires once and is gone, and the count still has to live somewhere persistent.

Change:

  • SidePanelLayout gains an optional badge: { count, label } on a tab. It renders a count pill (same look as the main rail's badge) in both the desktop rail and the mobile list, nothing at a count of 0, with role="status" and the label as its accessible name so a screen reader hears what the number counts.
  • CapabilitiesPage reads the pending queue with the same query key and fetcher as the Skills tab (['skills-pending']), so the two share one cache entry and the existing skills.pending_changed live frame refreshes both.
  • When that read fails, the rail footer shows an ErrorNotice ("Couldn't check for skills awaiting review", with the failure underneath) instead of a silent 0 that would read as "nothing is waiting". The notice has no Ask-the-agent hand-off: the footer sits beside whichever tab is open, the hand-off navigates to chat, and the Skills editor keeps an in-progress edit only in local state with no leave guard, so the click would discard that draft.
  • SkillsTab marks a candidate that adds a skill with a "New" Badge (the shared status primitive); update proposals keep their "Update" badge.
  • New strings skills_awaiting_review, skills_awaiting_review_failed and new_candidate in every locale catalog, with context notes for translators; the pseudolocale is regenerated.
  • The skills module spec notes where the queue is surfaced.

Backwards compatibility

Compatible: badge is a new optional field on SidePanelTab; every existing tab omits it and renders as before.

Removes nothing: the diff only adds a pill, a badge, a failure notice, three strings and one read of an endpoint the page already calls.

Tests

  • CapabilitiesPage.pendingSkillsBadge.test.tsx (new): with two candidates waiting and the Crewmates tab open, the Skills row carries a pill reading 2 named "Skills awaiting review (2)"; an empty queue shows no pill and no failure notice; a failed read shows no pill and shows the failure notice with the error text and no hand-off.
  • SkillsTabPendingUpdates.test.tsx: a plain new candidate is marked New; an update candidate is not.
  • Both files fail without the source change and pass with it.
  • Locally on Node 24: the i18n suites plus every SidePanelLayout, SkillsTab and CapabilitiesPage-rendering test file (77 files, 1161 tests) pass; tsc, eslint on the touched files, i18n-check, lint:i18n, lint:theme-colors, lint:phantom-classes and lint:restyle-ratchet are clean; npm run gen:ui produces no diff from this change.

Manual verification

Rendered the real CapabilitiesPage in a headless browser against stubbed API responses (two candidates: one new, one update; and a 503 from the queue endpoint), before and after the change, in dark and light themes. Screenshots below.

Screenshots / video

Another tab open (Crewmates), two candidates waiting. Before: nothing on the Skills row. After: a count pill.

Before After
Before: Crewmates tab, no count on Skills After: Crewmates tab, Skills row shows 2

Skills tab. Before: the new candidate has no badge. After: New beside Update, and the row keeps its count.

Before After
Before: Skills tab, new candidate unmarked After: Skills tab, new candidate marked New

The queue cannot be read (the endpoint answers 503): the rail footer says the check failed, and no count is shown.

After: queue read failed, failure notice in the rail footer

Light theme

Before: Skills tab, light theme

After: Skills tab, light theme

Related Issues

Closes #5643

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

Contributions are accepted under the project's Apache-2.0 license; see the Licensing section of CONTRIBUTING.md.

@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 10, 2026
@github-actions

github-actions Bot commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — 🟡 CONCERNS

Premise-level review of 651264f8d110e22baa8ac6e62f3f5e078b378ff6 — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

First-Principles-Verdict: CONCERNS

The rail-footer failure notice adds a page-wide error surface the Goal never asked for; without it, a failed read just shows no pill, exactly as today.

Not justified as shipped

  • Item 3 — oversized: the Goal is "can see that generated skills are waiting", but the diff adds a page-wide ErrorNotice in footer={pendingSkillsNotice} on every tab whenever a background count read fails. The zero option is no pill, which is how the page behaves on base. No spec or invariant requires it. I searched docs/system-specs/common/error-handling.md and website/AGENTS.md for a no-silent-zero rule and found none.
    Subtraction: delete pendingSkillsNotice, the footer= prop, the skills_awaiting_review_failed key in 14 catalogs and its context note, and the two footer test cases. The Skills tab's own queue still reports its failure when opened.
    Clears when: the mechanism is removed, or a human amends the Goal.
  • Item 2 — inherited: "an update proposal carries an 'Update' badge, but a candidate that would add a new skill carries no marker" is a symmetry argument. With the Update badge on update rows, an unbadged row already tells new from update. The change also rewrites feat(skills): stage similar auto-skills as versioned pending updates + notify on review #782's pin "still renders a plain new candidate as raw SKILL.md, with no badge".
    Subtraction: delete the !isUpdate Badge and the new_candidate key in 14 catalogs.
    Clears when: Surface auto-generated skills in the dashboard Skills panel with an enable prompt #5643 or a user report shows people misreading unbadged candidates.

What this change ships

Inventory (6 items) — 4 justified

Intent: let a user notice generated skills waiting for approval without opening the Skills tab. This is an ADDITION.

  1. The Skills row on the Agent Capabilities rail shows a count pill of waiting candidates from any tab — justified
  2. Each new candidate in the queue gets a "New" badge — inherited (mirrors the Update badge)
  3. When the queue can't be read, the rail footer shows "Couldn't check for skills awaiting review" on every tab — oversized
  4. SidePanelTab gets an optional badge field — justified
  5. Two new locale strings, skills_awaiting_review and new_candidate — justified
  6. The skills spec line now says where the queue is surfaced — justified

[FIRST-PRINCIPLES-REVIEWED] 651264f

Description read

[DESCRIPTION-READ] f7795dc47872ad17e3be379f6167e6df92759635c368971a6c37077e9e075b6d

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 651264f8d110e22baa8ac6e62f3f5e078b378ff6 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Check that a failing /skills/pending read doesn't leave the error notice in the rail footer on every tab when generated skills are switched off (the default).

[DESIGN-REVIEWED] 651264f

Description read

[DESCRIPTION-READ] 495d156f5085445b97ee7a4aad12e6938d3254094815c599f4c8adf695133d3b

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 10, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — 🟡 CONCERNS

UX-level review of 651264f8d110e22baa8ac6e62f3f5e078b378ff6 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: CONCERNS

The blind reader only guessed what the rail's "2" counts, and the failure notice shows the raw error "gateway unavailable" with nothing the user can do.

Watch

  • Rail pill read as a guess. The pill is the PR's main control. The blind reader saw it as "a count of something that needs my attention under Skills" but rated that "a guess". They could not tell whether it counts "the waiting items or the skills I already have", because both "Pending review (2)" and "Skills (2)" read 2 (shot-04, shot-07). The pill's meaning is only in title/aria-label, so sighted users never see "awaiting review". This hits every user who sees the pill, every time. It is friction, not failure: the Skills page itself makes the count clear. Fix: show the pill in the review-queue style, or re-shoot with different queued and installed counts so the read can be checked.
  • Raw error text in the failure notice. message={pendingSkillsError.message} renders "gateway unavailable" under the title (shot-05). The reader said "'gateway' means nothing to me" and expected clicking might retry, but nothing does. The notice says what happened but not what to do next. It is rare, but it shows on every tab for as long as the outage lasts. Fix: swap the raw message for one line of next step, such as "Will retry automatically" if the query refetches, or add a retry button.

Suggestions

  • Use the "New" badge's look for the existing "Update" badge too, or the reverse. Right now one uses Badge variant="ok" and the other a hand-rolled bg-accent-subtle span, so the two siblings look different.

[UX-REVIEWED] 651264f

Description read

[DESCRIPTION-READ] 4ff5bec60f0f339f7995142f85b8a81299aaf8cfa5d9cce51673523ad2a40f40

That sha256 names everything this verdict read: the PR description, plus the 10 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 10, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 651264f8d110e22baa8ac6e62f3f5e078b378ff6 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 651264f

Verdict parsed from the review's SHA-scoped output markers for commit 651264f8d110e22baa8ac6e62f3f5e078b378ff6.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 651264f8d110e22baa8ac6e62f3f5e078b378ff6: <one-sentence reason>

@github-actions

github-actions Bot commented Oct 10, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 651264f8d110e22baa8ac6e62f3f5e078b378ff6 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 651264f

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 651264f8d110e22baa8ac6e62f3f5e078b378ff6: <one-sentence reason>

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 10, 2026
@cruisercohen
cruisercohen force-pushed the feat/5643-surface-pending-skills branch from 14e393c to f939375 Compare October 10, 2026 23:43
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 10, 2026
@cruisercohen
cruisercohen force-pushed the feat/5643-surface-pending-skills branch from f939375 to f1d495f Compare October 10, 2026 23:58
@cruisercohen
cruisercohen marked this pull request as ready for review October 11, 2026 00:07
@cruisercohen
cruisercohen requested review from a team and KiroCrewCaptain as code owners October 11, 2026 00:07
@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 11, 2026
@cruisercohen
cruisercohen enabled auto-merge (squash) October 11, 2026 00:07
@cruisercohen
cruisercohen force-pushed the feat/5643-surface-pending-skills branch from f1d495f to 698a4e8 Compare October 11, 2026 00:23
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention readiness: passed Eligible automated validation passed for the current revision and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 11, 2026
iamwhatever
iamwhatever previously approved these changes Oct 11, 2026

@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 adds a count pill to the Agent Capabilities rail's Skills row showing how many generated skill candidates wait for approval, so they can be seen from any tab. It also adds a New badge beside the existing Update badge in the review queue. That solves the real gap: the queue only renders while the Skills tab is open, and the page often opens on another tab. I read the whole diff. The page uses the same ['skills-pending'] query key and api.skillsPending() fetcher as SkillsTab, so both read one cache entry, and the existing skills.pending_changed WebSocket invalidation refreshes both. The pill has the "Skills awaiting review (N)" accessible name and title, renders nothing at 0, and does not move the update dot. The backend handler catches its own errors and returns an empty list, so the red footer notice only shows on a transport failure, not in any normal configuration. Tests cover the pill from another tab, the zero case, the failure notice and the success case. Before/after screenshots are attached, all latest checks are green, and GPT/Opus/Design/First Principles have no findings. UX's CONCERNS (a bare number, and a failure notice with a raw error and no retry) are fair polish but have no disposition yet. I suggest the author answer them or file a follow-up, but they are advisory, not a defect. — crew-pr-reviewer (deep re-check)

@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 11, 2026
@cruisercohen
cruisercohen force-pushed the feat/5643-surface-pending-skills branch from 698a4e8 to 658534b Compare October 11, 2026 07:33
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: passed Eligible automated validation passed for the current revision merge conflict Branch has merge conflicts with its base — author must resolve before merge labels Oct 11, 2026
Generated skill candidates wait in the Skills tab's review queue, which renders only while that tab is open. The Agent Capabilities rail's Skills row now carries a count pill of waiting candidates (accessible name says what it counts), and a candidate that adds a new skill is badged New beside the existing Update badge.

Closes #5643
@cruisercohen
cruisercohen force-pushed the feat/5643-surface-pending-skills branch from 658534b to 651264f Compare October 11, 2026 07:45
@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 11, 2026
@cruisercohen

Copy link
Copy Markdown
Contributor Author

Rebased on main. The ui-index regen commit was dropped because main now carries the same regen. One test-only change goes beyond the conflict: the new findByText('New') wait in SkillsTabPendingUpdates.test.tsx now has an explicit 5 s timeout and a comment naming the skills then skills-pending query chain. The GPT review lane asked for it on the rebased head. No production code changed.

@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 shows how many generated skill candidates wait for approval as a count pill on the Agent Capabilities rail's Skills row, visible from any tab, and adds a New badge beside the existing Update badge in the review queue, which solves the real gap that the queue only renders while the Skills tab is open. I read the whole diff on head 651264f: CapabilitiesPage uses the same ['skills-pending'] key and api.skillsPending() fetcher as SkillsTab, so both read one cache entry and the existing invalidations (approve, dismiss, the skills.pending_changed frame) refresh the pill too; the pill renders nothing at 0, carries the "Skills awaiting review (N)" accessible name, and does not disturb the update dot on either the desktop rail or the mobile list. The rebase since our dismissed review changed only a test wait (an explicit 5 s bound on findByText('New') for the two chained queries), no production code. Tests cover the pill from another tab, the zero case, the failure notice and the New badge, all checks are green, and no change request is open. The UX CONCERNS (a bare number whose meaning lives only in title/aria-label, and the raw error text in the footer notice) and the First Principles subtraction of the footer notice are reasonable polish with no disposition yet; I suggest the author answer them here or in a follow-up, but none is a defect. — crew-pr-reviewer (deep re-check)

@cruisercohen
cruisercohen merged commit ab54185 into main Oct 11, 2026
105 of 108 checks passed
@cruisercohen
cruisercohen deleted the feat/5643-surface-pending-skills branch October 11, 2026 17:35
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 11, 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.

Surface auto-generated skills in the dashboard Skills panel with an enable prompt

2 participants