Skip to content

revert: drop the git working-tree badge from the composer footer - #12273

Merged
bolichen97 merged 1 commit into
mainfrom
refactor/remove-composer-git-badge
Sep 20, 2026
Merged

bolichen97 merged 1 commit into
mainfrom
refactor/remove-composer-git-badge

Conversation

@bolichen97

Copy link
Copy Markdown
Contributor

Problem / Motivation

The message box's bottom row carried a small git readout beside the branch
name: a changed-file count in a yellow pill, then ↑1 ↓2 for commits ahead and
behind. It was not a button. Clicking it did nothing. It said a shorter version
of what the Git panel already says in full, one click away in the sidebar.

Why it matters

A number beside the branch label reads as a fact about your working tree. This
one was fetched every 60 seconds, so between fetches it was a guess. The panel
it copied is the same data, complete, with the file names. Two places showing
one truth means one of them is wrong at any moment, and the smaller one is the
one with no file names to check it against.

It also cost the row its labels on a phone. At 390px the badge pushed the three
chips beside it down to k…, d… and m…; without it the same row reads
kiro…, demo-s…, main…. The badge's own comment said it stayed at narrow
widths on purpose, because it was then the only tree-state signal. The Git panel
is that signal, and it was before this badge existed.

What changed (motivation → approach → change)

Gone Where
The badge itself, its tooltip text, its four props website/src/components/ChatInput.tsx
The query that fed it, on a 60s interval website/src/pages/ChatPage.tsx
Its test file and its screenshot harness website/src/test/, website/scripts/

The badge was one <span role="status"> plus a useMemo that built its
tooltip. Both are deleted, along with projectGitDirty,
projectGitDirtyTruncated, projectGitAhead and projectGitBehind, and the
now-unused FileDiff icon import.

ChatPage fed it from a ['git-status', dir] query on a 60-second interval,
plus an effect that refetched when the agent stopped running. Nothing else read
that query's result, so it goes too. The Git panel owns its own fetch on the
same key, so the panel loses nothing. A project with the panel closed now makes
no git status request at all.

Two files mentioned the badge and had to stop. The dashboard spec said the
footer badge was the silent one of three ['git-status', …] consumers; there
are two consumers now, and the sentence says so. A sibling capture script
pointed readers at the harness this change deletes, so that pointer is gone.

The Git panel, the branch label, the folder chip and the context readout are
all untouched.

Backwards compatibility

Compatible: nothing accepts less than before. The removed props were internal
to two components in this repo, the API routes are unchanged, and no setting
turned the badge on or off.

Tests

No new tests. The change removes a component and its only test file
(ChatInput.gitBadge.test.tsx, which asserted the badge's segments, tooltip and
hidden-when-clean behaviour). The frontend suite's other ChatInput tests and
the GitPanel tests run unchanged and cover what is left.

Manual verification

Both frames below come from the badge's own screenshot harness, run against the
real built dashboard with the git endpoints fixtured to 3 uncommitted files, 1
ahead and 2 behind — the exact input the badge existed to render. On this branch
the harness finds zero elements matching the badge, in dark, in light, and at
390px. The main frames are the same harness on origin/main, which finds one.

Screenshots / video

Before, on main — mainline · [3] ↑1 ↓2:

Composer footer on main: branch name followed by a yellow 3 pill and up/down arrow counts

After, on this branch — the branch name, nothing after it:

Composer footer on this branch: branch name with no git badge

At 390px the badge was taking the room the labels needed. Before, then after:

Narrow footer on main: chips truncated to single letters beside the badge

Narrow footer on this branch: the same chips readable

Light theme, and the full composer with the Git panel open

Light theme footer before

Light theme footer after

Git panel open on this branch, showing the per-file status the badge summarised

Related Issues

no linked issue: removal requested directly by the maintainer.

Pattern harvest

Rule candidate: review-prompt
Pattern: a second, smaller copy of an existing panel's data, added as a passive
readout with no action, which restates the panel and goes stale on its own
interval.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|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

@bolichen97
bolichen97 requested review from a team and cixuuz September 20, 2026 10:32
@bolichen97

Copy link
Copy Markdown
Contributor Author

Intent: Remove the composer footer's git working-tree badge, so the Git panel is the one place that reports working-tree state.
Not a goal: Changing the Git panel, the branch label, the /api/project/git/status route, or any other footer chip.

@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Sep 20, 2026
@github-actions

github-actions Bot commented Sep 20, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Fable 5) — ✅ PASS

Design-level review of 3506e35a559cefce718c5b04e825850d95e16c6c — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Removal cites "requested directly by the maintainer" with no linked issue — confirm that authorization exists before merging a feature deletion.

[DESIGN-REVIEWED] 3506e35

@github-actions

github-actions Bot commented Sep 20, 2026 •

Copy link
Copy Markdown
Contributor

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of 3506e35a559cefce718c5b04e825850d95e16c6c and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 3506e35

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 3506e35a559cefce718c5b04e825850d95e16c6c: <one-sentence reason>

@github-actions

github-actions Bot commented Sep 20, 2026 •

Copy link
Copy Markdown
Contributor

Opus 4.8 Review — ✅ no blocking findings

Reviewed 3506e35a559cefce718c5b04e825850d95e16c6c — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 3506e35

Verdict parsed from the review's SHA-scoped output markers for commit 3506e35a559cefce718c5b04e825850d95e16c6c.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 3506e35a559cefce718c5b04e825850d95e16c6c: <one-sentence reason>

@github-actions

github-actions Bot commented Sep 20, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Fable 5) — ✅ human override accepted

@bolichen97 overrode this lane for 3506e35a559cefce718c5b04e825850d95e16c6c via /ai-review override (this commit only).

@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 Sep 20, 2026
@github-actions

github-actions Bot commented Sep 20, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Fable 5) — ✅ PASS

UX-level review of 3506e35a559cefce718c5b04e825850d95e16c6c — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: PASS

Confirm the Git panel is still reachable at 390px, since after removal it is the only working-tree signal on narrow viewports.

[UX-REVIEWED] 3506e35

@bolichen97

Copy link
Copy Markdown
Contributor Author

self-added: no

  • needs-a-decision: removal of a shipped, always-on surface with no RFC covering it span=ee3ffa6e3406

The lane is factually right on every point. The badge shipped on for everyone, no setting gated it, and the base RFC list carries nothing covering the composer footer or this badge. I am not rebutting it and I am not filing an issue for it: the lane's own Clears when: line says the only path is a maintainer decision, so this is put to the maintainer rather than answered in code.
The provenance the lane cannot see: the repository owner looked at the badge in the running dashboard and asked for it to be deleted. That is the authority, and only he can record it.
Scope of the removal, so the decision is made on facts: the composer-footer badge only (changed-file pill plus the ahead/behind counts), its tooltip, its four props, the 60-second poll that fed it, its own test file and capture script. The Git panel is untouched and still shows the same data in more detail, so no information is lost from the product.
@bolichen97 — to clear this lane, post as its own top-level comment: /ai-review override first-principles c18bbbee028e8dc1d2b1a336c069940c15d29ded: <your reason>

@bolichen97

Copy link
Copy Markdown
Contributor Author

self-added: no

  • needs-a-decision: the BLOCKERS item restating the same missing-RFC premise span=32c926ca57be

Same premise as the lane's other item, verified separately against the same evidence: the 51-entry base RFC status list names no RFC for the composer footer or this badge, and the description's only cited authority is the maintainer's direct request. The lane is correct that a PR description is not an accepted design record.
Not rebutted and not deferred. A deferral would be dishonest here, because there is no later task to pick up: either the owner rules that his direct request is sufficient authority for deleting one always-on badge, or the change should not land at all. There is no third outcome I can execute.
I am deliberately not writing an RFC to satisfy the gate. An RFC for a one-badge deletion whose replacement surface already exists would be paperwork produced to clear a check, not a design record anyone would read.
@bolichen97 — the decision, and the override that records it, are yours: /ai-review override first-principles c18bbbee028e8dc1d2b1a336c069940c15d29ded: <your reason>

@bolichen97

Copy link
Copy Markdown
Contributor Author

Backend Tests (Windows) (1) is base debt, not this diff.

Four test_acp_runtime.py kill-summary tests assert [returncode=-15], a POSIX SIGTERM value Windows cannot produce. Already tracked as #12247, with three open fixes: #12271, #12269, #12260.

Evidence this PR cannot be the cause:

  • git -P diff --name-only origin/main...HEAD is six files, none of them Python: docs/system-specs/modules/learn-cron-dashboard.md, website/scripts/capture-footer-git-status.mjs, website/scripts/capture-git-panel-degraded-states.mjs, website/src/components/ChatInput.tsx, website/src/pages/ChatPage.tsx, website/src/test/ChatInput.gitBadge.test.tsx.
  • The failing tests arrived with fix(acp): say who killed a runtime and never log a bare returncode=None #12188 (fix(acp): say who killed a runtime and never log a bare returncode=None), which is an ancestor of this head's base.
  • main's own push runs do not schedule that shard, so main never exercised it and is red the same way.

No rerun: the shard cannot pass on any head until one of the three fixes merges. Not folding an unrelated ACP fix into a frontend deletion.

The composer footer carried a second, smaller copy of the Git panel: a
changed-file count and ahead/behind arrows beside the branch name. It was
a passive readout with no action, it repeated what the Git panel already
shows in full, and it put a number next to the branch label that reads as
a claim about the tree the moment it goes stale.

Removing it takes out the badge markup, its tooltip builder and its four
props from ChatInput, plus the ChatPage query that fed it: a
['git-status', dir] observer on a 60s interval with a running-to-idle
invalidation, which existed only for the badge. The Git panel keeps its
own fetch on the same key, so nothing else loses data, and a project
with no panel open no longer polls git status at all.

The badge's own test file and its screenshot harness go with it. The
dashboard spec's sentence about the footer badge being the silent
['git-status', ...] consumer is rewritten to the two consumers that
remain, and the sibling capture script no longer points at a script that
does not exist.
@bolichen97
bolichen97 force-pushed the refactor/remove-composer-git-badge branch from c18bbbe to 3506e35 Compare September 20, 2026 11:54
@bolichen97

Copy link
Copy Markdown
Contributor Author

Rebased: main moved from 4dfbfb04 to 954fb691 and now carries #12269 (fix(acp): record the reap on the Windows teardown branch too), which is the fix for the Backend Tests (Windows) (1) red this PR was stuck behind (#12247). CI re-running on 3506e35a.

The patch itself is unchanged -- git range-diff origin/main c18bbbee02 3506e35a55 reports =, and the diff is the same six files. Re-verified after the rebase: tsc -p tsconfig.app.json clean, ChatInput.test.tsx 163 passed.

Also on the previous head, Backend Tests (3.12, 8) went red twice for a reason that is not this diff: the xdist worker was killed at ~91% through the shard, once as gw3 and once as gw2, and the run's own budget line reads xdist worker budget: 4 of 8 workers (14.0 GiB free, 15 GiB installed). Each worker needs about 3 GiB. The test named in the failure (test_source_corpus.py::TestNoGateEnumeratesTheCheckoutByWalking::test_no_gate_scans_the_checkout_from_the_filesystem) is simply the one in flight when the worker died; it passes locally on this head, and the same shard is green on #12271 and #12260. Coverage Gate was a fail-closed consequence of it (backend-test=failure -- failing closed), not an independent finding.

Still blocked on the First Principles lane, which needs a maintainer override on the new head sha.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Sep 20, 2026
@bolichen97

Copy link
Copy Markdown
Contributor Author
  • needs-a-decision: product-shape change without an accepted RFC span=772e5de45c07

Re-posted for head 3506e35a559c after the rebase onto 954fb691; the patch is unchanged (git range-diff reports =, same six files).

The lane is right about the record, and I am not rebutting it. This PR deletes a shipped, default-on, user-visible readout, and the only provenance on record is the maintainer asking for it directly. No RFC covering the composer-footer git badge is non-draft on the base, so the gate has nothing to read.

I cannot clear this myself. Either path belongs to a human: merge an RFC on the base, or record /ai-review override first-principles 3506e35a559cefce718c5b04e825850d95e16c6c: <decision>. This PR is same-repo, so the override lane is available.

What is verifiable from the diff, for whoever decides: the removed surface is display-only (a changed-file count pill plus ↑ahead ↓behind) and its data is still shown, in more detail, in the Git panel this PR does not touch. Nothing else reads the deleted props, and the 60-second polling query that fed only this badge is deleted with it.

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • needs-a-decision: deleted contract test named in the BLOCKERS item span=f74abb35c9b1

Re-posted for head 3506e35a559c after the rebase onto 954fb691; the patch is unchanged (git range-diff reports =).

The observation is accurate: the deleted test file's header reads "Pins the footer git-status badge contract", so removing it removes the pin for the surface being removed. That is the intended shape of a revert, not an accidental loss of coverage: the test asserted the badge renders its counts, and there is no badge left to assert. Keeping the file would leave a test with no subject.

Coverage that remains and was re-run on this head: npx tsc -p tsconfig.app.json reports 0 errors, and src/test/ChatInput.test.tsx passes 163 cases over the component that used to host the badge. Nothing in the repo still references the removed props, the removed query key, or the deleted tooltip.

This does not clear on its own either, for the same reason as span 772e5de45c07: the gate wants an RFC on the base or a maintainer override at 3506e35a559cefce718c5b04e825850d95e16c6c. Only a human can post it.

@bolichen97

Copy link
Copy Markdown
Contributor Author

/ai-review override first-principles 3506e35: maintainer decision: the badge duplicates the Git panel and crushes the session labels at 390px

@bolichen97
bolichen97 enabled auto-merge (squash) September 20, 2026 18:11
@github-actions

Copy link
Copy Markdown
Contributor

Human judgment recorded

@bolichen97 marked the first-principles AI finding as false positive, not applicable, or explicitly accepted for 3506e35a559cefce718c5b04e825850d95e16c6c.

maintainer decision: the badge duplicates the Git panel and crushes the session labels at 390px

This decision applies only to this commit. A new push requires a new judgment.

@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 Sep 20, 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 Sep 20, 2026
@bolichen97
bolichen97 merged commit 8169d3c into main Sep 20, 2026
93 of 94 checks passed
@bolichen97
bolichen97 deleted the refactor/remove-composer-git-badge branch September 20, 2026 18:22
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Sep 20, 2026
amadsalmon pushed a commit to amadsalmon/KiroCrew that referenced this pull request Sep 22, 2026
…odotdev#12273)

The composer footer carried a second, smaller copy of the Git panel: a
changed-file count and ahead/behind arrows beside the branch name. It was
a passive readout with no action, it repeated what the Git panel already
shows in full, and it put a number next to the branch label that reads as
a claim about the tree the moment it goes stale.

Removing it takes out the badge markup, its tooltip builder and its four
props from ChatInput, plus the ChatPage query that fed it: a
['git-status', dir] observer on a 60s interval with a running-to-idle
invalidation, which existed only for the badge. The Git panel keeps its
own fetch on the same key, so nothing else loses data, and a project
with no panel open no longer polls git status at all.

The badge's own test file and its screenshot harness go with it. The
dashboard spec's sentence about the footer badge being the silent
['git-status', ...] consumer is rewritten to the two consumers that
remain, and the sibling capture script no longer points at a script that
does not exist.
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