[juliusmarminge]

Nothing needed changing this round. PR #36938 at f6260a4: all 8 required checks passed, Macroscope approved it (review and Approvability verdict), there are no line comments, no failing or pending check runs, and no high-or-above bot findings. The github-actions thread-transfer comment reports that the PR stays within every ceiling. The fix is unchanged from the previous round: the Conventional Commits preset now asks for a lowercase description that does not open with a proper noun or acronym, and GitManager.sanitizeCommitMessage lowercases a capitalized first word as a backstop. The watch is stopped.

Fixes #36920

Reproduction on main

Commit a20c4db (origin/main), web client against T3CODE_DEV_SEED=0 vp run dev (fresh database). Test repo: three lowercase conventional commits, @commitlint/cli + @commitlint/config-conventional run from a commit-msg hook (core.hooksPath). Unstaged change: .github/workflows/e2e.yml trigger from push: main to pull_request + workflow_dispatch. Settings: Source control > Commit and pull request writing = Conventional Commits; text generation model = Codex gpt-6-astra. Added the folder as a project, clicked Commit in the thread's Version Control panel with no message (the app generates one), then git reset --soft HEAD~1 && git reset and repeated. 1 of 5 gpt-6-astra runs generated ci: Run E2E on pull requests and manual dispatch instead of pushes. commitlint rejected it (subject must not be sentence-case [subject-case], exit 1), and the app showed the toast 'Action failed: Git command failed in GitVcsDriver.commit.commit (): Git command exited with a non-zero status.' with no hook output, and the change left staged. The other 4 Astra runs gave ci: run E2E on pull requests and manual dispatch and passed; 5 runs on the default gpt-6-luna all passed (lowercase). Same prompt built with buildCommitMessagePrompt + the conventional_commits policy and sent straight to the models via OpenCode: gpt-6-astra capitalized 4/5 (ci: Run E2E on pull requests and manual dispatch instead of pushes x3, ci(e2e): Run on pull requests and allow manual dispatch); gpt-6-luna 0/8, claude-haiku-5-5 0/8, claude-opus-4-8 0/5, big-pickle 0/5, glm-5.3-flash 0/5.

Evidence

Reproduced in the real web app with fresh databases, Chromium at 1280×800, light theme, and commitlint's conventional config. Seeded three lowercase conventional commits, changed the E2E triggers from pushes to pull requests and manual dispatch, and clicked Commit with no message. Replayed the exact reported capitalized model response at the text-generation CLI boundary on both sides.

scenarios:prove exited 0: failed on base and passed on head.

Evidence	Before — f56f27e6c9	After — f6260a4f2a
Result	ci: Run E2E …; commitlint rejected subject-case, exit 1; generic failure toast	ci: run E2E …; commitlint exited 0; commit created
Scenario capture
Recorded success
Interaction	Before recording	After recording
Hook transcript	Rejected capitalized subject	Accepted lowercase subject

Scenario source · Proof summary · Commitlint dependencies · Dependency lockfile

Synthetic repository paths were redacted. Live Astra sampling was not performed; this proves the app's handling of the reported model response.

Not shown:

• The new prompt instruction's effect on live GPT-6-Astra output was not measured; the scenario replayed the exact reported capitalized response to verify the sanitizer and real commit flow deterministically.

Tests

No new commands this round; CI on f6260a4 passed all 8 required checks. The previous round ran vp run check:changed (format, lint, knip, typecheck, tests all passed) and a live gpt-6-astra check: old prompt 2/5 capitalized, new prompt 0/5.

Not in this change

• When a commit-msg hook rejects a commit, GitVcsDriverCore replaces the hook's output with the generic 'Git command exited with a non-zero status.', so the user cannot see why. Surfacing hook output is a separate change (it touches the deliberate rule that only Git's own diagnostic lines are trusted).
• The conventional_commits preset prompt (TextGenerationPresets.ts) does not ask for a lowercase description. Adding that changes model-facing text, which needs the evals rerun; the deterministic normalization makes it unnecessary for correctness.
• provider-core sanitizeCommitSubject applies no style-aware normalization; GitManager's sanitizer already covers every commit path that uses a writing style.
• When a commit-msg hook rejects a commit, GitVcsDriverCore still replaces the hook's output with 'Git command exited with a non-zero status.', so the user cannot see why. That is a separate change.
• There is no eval for generated commit messages; adding one under evals/ would let future changes to commit prompts be measured instead of spot-checked.

Who did the work

• Written by Claude Opus 5.5, in Claude Code
• Reviewed by GPT-6.1 Sol, in Codex, and by Claude Opus 5.5, each on its own
• Proof captured by GPT-6.1 Sol and checked by Claude Opus 5.5

Run by the t3code factory in T3 Code.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	21.5 KiB	21.5 KiB	0 B (0.0%)	24.4 KiB	✅
Codex	Live turn messages	1	1	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.9 KiB	21.9 KiB	0 B (0.0%)	24.4 KiB	✅
Claude	Live turn messages	1	1	0 (0.0%)	8	✅

Other scenarios

All 64 scenario measurements match main.

All 64 scenario measurements

Provider	Scenario	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	short turn resume	WebSocket decoded	10.5 KiB	10.5 KiB	0 B (0.0%)	12.7 KiB	✅
Codex	short turn resume	WebSocket messages	2	2	0 (0.0%)	2	✅
Codex	long turn running, cold open	HTTP decoded	281.7 KiB	281.7 KiB	0 B (0.0%)	302.7 KiB	✅
Codex	long turn running, cold open	rows	75	75	0 (0.0%)	75	✅
Codex	long turn running, history page	HTTP decoded	92.4 KiB	92.4 KiB	0 B (0.0%)	107.4 KiB	✅
Codex	long turn running, history page	rows	75	75	0 (0.0%)	75	✅
Codex	long turn finished, cold open	HTTP decoded	381.5 KiB	381.5 KiB	0 B (0.0%)	410.2 KiB	✅
Codex	long turn finished, cold open	rows	151	151	0 (0.0%)	151	✅
Codex	long turn finished, socket cold open	WebSocket decoded	381.6 KiB	381.6 KiB	0 B (0.0%)	410.2 KiB	✅
Codex	long turn finished, socket cold open	WebSocket messages	2	2	0 (0.0%)	2	✅
Codex	long turn finished, socket cold open	rows	151	151	0 (0.0%)	151	✅
Codex	long turn finished, work page	HTTP decoded	186.2 KiB	186.2 KiB	0 B (0.0%)	200.2 KiB	✅
Codex	long turn finished, work page	rows	150	150	0 (0.0%)	150	✅
Codex	idle thread resume	WebSocket decoded	65 B	65 B	0 B (0.0%)	200 B	✅
Codex	idle thread resume	WebSocket messages	1	1	0 (0.0%)	1	✅
Codex	fleet shell snapshot	HTTP decoded	42.4 KiB	42.4 KiB	0 B (0.0%)	53.7 KiB	✅
Codex	fleet shell snapshot	rows	27	27	0 (0.0%)	27	✅
Codex	fleet shell subscribe	WebSocket decoded	65 B	65 B	0 B (0.0%)	200 B	✅
Codex	fleet shell subscribe	WebSocket messages	1	1	0 (0.0%)	1	✅
Codex	fleet shell resume check	HTTP decoded	18 B	18 B	0 B (0.0%)	100 B	✅
Codex	fleet shell resume check	rows	0	0	0	0	✅
Codex	stale shell reload	HTTP decoded	42.4 KiB	42.4 KiB	0 B (0.0%)	53.7 KiB	✅
Codex	stale shell reload	rows	27	27	0 (0.0%)	27	✅
Codex	stale shell resume	WebSocket decoded	140 B	140 B	0 B (0.0%)	400 B	✅
Codex	stale shell resume	WebSocket messages	2	2	0 (0.0%)	2	✅
Codex	stale todo list cold open	HTTP decoded	72.4 KiB	72.4 KiB	0 B (0.0%)	83.0 KiB	✅
Codex	stale todo list cold open	rows	75	75	0 (0.0%)	75	✅
Codex	stale todo list cold open	nodes	2	2	0 (0.0%)	4	✅
Codex	stale todo list cold open	subagents	2	2	0 (0.0%)	4	✅
Codex	stale todo list cold open	plans	1	1	0 (0.0%)	2	✅
Codex	stale todo list cold open	providerTurns	1	1	0 (0.0%)	2	✅
Codex	stale todo list cold open	checkpoints	1	1	0 (0.0%)	2	✅
Claude	short turn resume	WebSocket decoded	10.6 KiB	10.6 KiB	0 B (0.0%)	12.7 KiB	✅
Claude	short turn resume	WebSocket messages	2	2	0 (0.0%)	2	✅
Claude	long turn running, cold open	HTTP decoded	283.3 KiB	283.3 KiB	0 B (0.0%)	302.7 KiB	✅
Claude	long turn running, cold open	rows	75	75	0 (0.0%)	75	✅
Claude	long turn running, history page	HTTP decoded	93.4 KiB	93.4 KiB	0 B (0.0%)	107.4 KiB	✅
Claude	long turn running, history page	rows	75	75	0 (0.0%)	75	✅
Claude	long turn finished, cold open	HTTP decoded	384.2 KiB	384.2 KiB	0 B (0.0%)	410.2 KiB	✅
Claude	long turn finished, cold open	rows	151	151	0 (0.0%)	151	✅
Claude	long turn finished, socket cold open	WebSocket decoded	384.3 KiB	384.3 KiB	0 B (0.0%)	410.2 KiB	✅
Claude	long turn finished, socket cold open	WebSocket messages	2	2	0 (0.0%)	2	✅
Claude	long turn finished, socket cold open	rows	151	151	0 (0.0%)	151	✅
Claude	long turn finished, work page	HTTP decoded	188.1 KiB	188.1 KiB	0 B (0.0%)	200.2 KiB	✅
Claude	long turn finished, work page	rows	150	150	0 (0.0%)	150	✅
Claude	idle thread resume	WebSocket decoded	66 B	66 B	0 B (0.0%)	200 B	✅
Claude	idle thread resume	WebSocket messages	1	1	0 (0.0%)	1	✅
Claude	fleet shell snapshot	HTTP decoded	43.0 KiB	43.0 KiB	0 B (0.0%)	53.7 KiB	✅
Claude	fleet shell snapshot	rows	27	27	0 (0.0%)	27	✅
Claude	fleet shell subscribe	WebSocket decoded	66 B	66 B	0 B (0.0%)	200 B	✅
Claude	fleet shell subscribe	WebSocket messages	1	1	0 (0.0%)	1	✅
Claude	fleet shell resume check	HTTP decoded	18 B	18 B	0 B (0.0%)	100 B	✅
Claude	fleet shell resume check	rows	0	0	0	0	✅
Claude	stale shell reload	HTTP decoded	43.0 KiB	43.0 KiB	0 B (0.0%)	53.7 KiB	✅
Claude	stale shell reload	rows	27	27	0 (0.0%)	27	✅
Claude	stale shell resume	WebSocket decoded	140 B	140 B	0 B (0.0%)	400 B	✅
Claude	stale shell resume	WebSocket messages	2	2	0 (0.0%)	2	✅
Claude	stale todo list cold open	HTTP decoded	73.4 KiB	73.4 KiB	0 B (0.0%)	83.0 KiB	✅
Claude	stale todo list cold open	rows	75	75	0 (0.0%)	75	✅
Claude	stale todo list cold open	nodes	2	2	0 (0.0%)	4	✅
Claude	stale todo list cold open	subagents	2	2	0 (0.0%)	4	✅
Claude	stale todo list cold open	plans	1	1	0 (0.0%)	2	✅
Claude	stale todo list cold open	providerTurns	1	1	0 (0.0%)	2	✅
Claude	stale todo list cold open	checkpoints	1	1	0 (0.0%)	2	✅

Baseline: 74f02db · PR result: f6260a4 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at f6260a4

Macroscope's review found this PR approvable — Generated Conventional Commit subjects now favor lowercase descriptions, with a narrowly scoped normalization backstop. Explicit commit messages and the default repository-conventions mode remain unchanged.

^{You can add or adjust custom eligibility rules. Learn more.}

[coderabbitai]

No actionable comments were generated in the recent review.

Walkthrough

Generated Conventional Commit messages now lowercase the first word of the description. The generation instructions specify this capitalization, and a test checks messages with types, scope, and a breaking-change marker.

Changes: Generated commit subjects

Layer / File(s)	Summary
Sanitize generated Conventional Commit subjects apps/server/src/git/GitManager.ts, apps/server/src/textGeneration/TextGenerationPresets.ts, apps/server/src/git/GitManager.test.ts	The sanitizer lowercases the first capitalized description word when the selected policy is Conventional Commits. Generated messages pass the policy to the sanitizer. The generation instructions and test specify lowercase descriptions.

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix · Severity of issue fixed: Low

Suggested reviewers: saphid, only21mil

Merge Risk: Minimal · up to f6260

Generated Conventional Commit subjects now start with a lowercase word, which reduces commitlint rejections. The change is small and limited to the Conventional Commits policy, and no merge-blocking risk was found.

Pre-merge checks: 4

Check name	Status	Explanation
Title check		The title names the main change: Conventional Commit subjects generated by the server now start with a lowercase description. It uses the repo's conventional commit format and fits the objectives, whi…
Linked Issues check		Issue #36920 requires generated commit messages to use valid Conventional Commit format. The reported failure is a capitalized description (ci: Run E2E ...) rejected by commitlint. The head adds two…
Out of Scope Changes check		The whole-PR diff against the merge base touches three files. TextGenerationPresets.ts adds one lowercase-description instruction to the Conventional Commits preset. GitManager.ts changes `sanitiz…
Description check		The description leads with the outcome of the PR and includes a reproduction on main, a before/after verification table with evidence, and a list of out-of-scope items. It also names the linked issue …

Full details: Title check

Explanation

The title names the main change: Conventional Commit subjects generated by the server now start with a lowercase description. It uses the repo's conventional commit format and fits the objectives, which target commitlint rejections of capitalized descriptions. The title is specific and concise.

Full details: Linked Issues check

Explanation

Issue #36920 requires generated commit messages to use valid Conventional Commit format. The reported failure is a capitalized description (ci: Run E2E ...) rejected by commitlint. The head adds two mechanisms. (1) In TextGenerationPresets.ts, conventionalCommitsTextGenerationPolicy.commitInstructions now tells the model to start the description with a lowercase word. (2) In GitManager.ts, sanitizeCommitMessage takes a conventionalCommits flag. When the resolved policy is conventional_commits, the anchored regex CONVENTIONAL_CAPITALIZED_DESCRIPTION lowercases a capitalized first word after the type(scope)!: prefix. The regex leaves acronyms and mixed-case tokens such as E2E, API, and GitHub unchanged, because the lookahead rejects a following alphanumeric character. The diff also shows the GitManager.ts sanitizer change applied to the generated commit path. The PR description reports a test in GitManager.test.ts that covers capitalized subjects. I did not read that test file, so its exact assertions are not verified here. The code change addresses the issue requirement.

Full details: Out of Scope Changes check

Explanation

The whole-PR diff against the merge base touches three files. TextGenerationPresets.ts adds one lowercase-description instruction to the Conventional Commits preset. GitManager.ts changes sanitizeCommitMessage to take the policy flag and passes policy?.kind === "conventional_commits" at the generation call site. Its other changes are the regex constant and the comment. Both changes serve issue #36920 directly. The test file is described as covering the same behavior. The PR description lists hook-output visibility and commit-message evals as separate work. The diff contains neither change, so no unrelated change is present.

Full details: Description check

Explanation

The description leads with the outcome of the PR and includes a reproduction on main, a before/after verification table with evidence, and a list of out-of-scope items. It also names the linked issue (#36920). It does not follow the required template headings (Problem, Change, Scope, Verification), and the opening paragraph is largely a status report about checks and bot reviews rather than a description of the change. The Scope section (issue link and bug-fix justification) is present in substance, but the Problem and Change sections are not labeled or cleanly separated. The description also repeats the 'hook output hidden' item twice in the Not in this change section, and it includes a 'Who did the work' section the template does not ask for. Overall the content is substantive and on topic, so the check is mostly complete, but the structure does not match the template.

Finishing Touches

Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Recent review info

Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: factory/issue-36920
• File: .coderabbit.config.ts
• Commit: f6260a4
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Team
• Run ID: 303c362d-570b-4af2-a5bc-0ed553b4c1c4

Commits

Reviewing files that changed from the base of the PR and between 37c266d and f6260a4.

Files selected for processing: 3

• apps/server/src/git/GitManager.test.ts
• apps/server/src/git/GitManager.ts
• apps/server/src/textGeneration/TextGenerationPresets.ts

Limit details: You’ve used all 10 included reviews currently available.