fix(governance-paths): recognize trusted-bots.yml and main.json as governance-sensitive

[tvna]

Summary

Add .github/trusted-bots.yml and .github/rulesets/main.json to _GOVERNANCE_FILENAMES in gitapex_check_canonical_governance_paths.py so both are correctly classified as governance-sensitive files.

Facts

• _GOVERNANCE_FILENAMES (as of original origin/main, lines 64-78) did not include .github/trusted-bots.yml.
• classify() only matches governance via normalized in _GOVERNANCE_FILENAMES, so .github/trusted-bots.yml did not match any prefix rule either and fell through to no-match.
• .github/trusted-bots.yml was added by PR feat(ci): exempt trusted bots from the independent-review-pending gate #1859 (issue feat(ci): exempt trusted bots from independent-review-pending's human-verdict requirement #1858): an identity-based bypass allowlist for the independent-review-pending merge gate.
• The corresponding test file test_gitapex_check_canonical_governance_paths.py (same directory) had no case for trusted-bots.yml; 13 existing tests confirmed.
• Step 8's independent adversarial review (see Execution log) found that .github/CODEOWNERS documents .github/rulesets/main.json's commit_author_email_pattern/committer_email_pattern rules as "the same class of risk" as trusted-bots.yml (issue feat(ci): exempt trusted bots from independent-review-pending's human-verdict requirement #1858's second trust anchor), yet this classifier left main.json uncovered (no-match) too. Repository owner approved widening this PR to cover it as well.

Assumptions

None (implementation follows issue #1875's own Acceptance Criteria Map, amended per the Step 8 finding above with owner approval -- see issue #1875's own "Scope amendment" section).

Risk / blast radius

Low. Adds two entries to the _GOVERNANCE_FILENAMES tuple plus regression tests; no other classification list or branch (_WORKFLOW_PREFIXES / _HOOK_SCRIPT_PREFIXES / _DEPENDENCY_MANIFEST_FILENAMES, etc.) is touched. This classifier is a deterministic pre-filter only; both .github/trusted-bots.yml and .github/rulesets/main.json already carry a CODEOWNERS entry requiring @tvna's review, so this gap was not itself an exploitable vulnerability before this fix -- only a completeness gap in one defense-in-depth layer.

Rollback

Revert these commits to restore _GOVERNANCE_FILENAMES to its prior state. No other protection (CODEOWNERS, etc.) is touched, so it remains in place after a revert.

Verification

Acceptance Criteria Map (carried over from issue #1875, amended per its own "Scope amendment" section; a draft, independently re-check each row against this PR's own diff rather than trusting it as-is):

Criterion	Interpretation	Planned ops	Proof method	Result
Classifier recognizes .github/trusted-bots.yml as governance	Add .github/trusted-bots.yml as one entry in the _GOVERNANCE_FILENAMES tuple	Add one line (with a rationale comment) to _GOVERNANCE_FILENAMES	Add regression test; pytest passes in full	Done. Committed at 36563c94. Manual CLI check confirms governance: .github/trusted-bots.yml
Classifier recognizes .github/rulesets/main.json as governance (Scope amendment)	Add .github/rulesets/main.json as one entry in the _GOVERNANCE_FILENAMES tuple	Add one line (with CODEOWNERS's own rationale as comment) to _GOVERNANCE_FILENAMES	Add regression test; pytest passes in full	Done. Committed at 1ebd9207. Manual CLI check confirms governance: .github/rulesets/main.json, and confirms no collision with .github/scripts/ (hook-script) classification
Existing governance-file classification logic stays intact	No other classification list/logic is touched	Limit the change to the tuple entries and their own regression tests	All 13 existing tests plus new tests pass	Done. Full repo suite and local-preflight (47 gates) also pass after every commit

Checklist

• Tests pass locally
• Docs updated if behavior changed
• Issue number cited in every commit
• If this PR adds/modifies a skills/*/SKILL.md, a docs/superpowers/specs/*.md design doc, a security-relevant skill, or a deterministic checker script (skills/*/scripts/*.py, evals/scripts/*.py, .github/scripts/*.py), a ## Skill audit evidence section discloses the required verdicts/waivers (see .github/scripts/gitapex_gate_skill_audit_disclosure.py)
• If this PR adds a new Kept-edit-log entry to any evals/*/split.md, that entry discloses a Transfer check line (see .github/scripts/gitapex_gate_transfer_check_disclosure.py)
• If this PR adds or increases a skills/*/SKILL.md's Stop-boundary bullets or named dispatch branches, evals/<skill>/tasks/*.yaml gained at least as many new fixtures (see .github/scripts/gitapex_gate_skill_branch_fixture_coverage.py)

Skill audit evidence

• checker-script-adversarial-review: RAN -- an independent adversarial review (Step 8, review-persona dispatch) checked path-normalization edge cases, interaction with _HOOK_SCRIPT_PREFIXES, near-miss path evasion, and rationale accuracy against the real consuming gate and CODEOWNERS for skills/executing-a-branch-plan/scripts/gitapex_check_canonical_governance_paths.py. It found one real completeness gap (the .github/rulesets/main.json finding described above), which this PR now also addresses. A separate behavior-preserving refactor/simplify pass (branch-plan-task dispatch) found the added lines already consistent with the file's existing conventions and made no changes.
• defeat-test-disclosure: RAN -- added tests confirming an extension/case near-miss variant of either new entry is NOT swept in as a false governance match, per the review's own finding above, and confirming .github/rulesets/main.json is not accidentally absorbed by the .github/scripts/-prefixed hook-script or .github/workflows/-prefixed workflow branches.

Independent review verdict

• Verdict: CLEAN
• Verified commit: 638982f

Outer layer: GitHub Copilot review was requested (request_copilot_review) shortly after the inner-layer fan-out below started. The repository owner directed this session to proceed without awaiting Copilot's response, so this layer is disclosed as requested-but-not-awaited, not as a clean pass. No confirmation that Anthropic's "Claude Code Review" GitHub App is installed on this repository.

Inner layer (reviewing-an-artifact, low effort, against commit 1ebd9207's diff -- trusted-bots.yml plus main.json additions and their tests):

• Classification: dangerous/mixed (the diff edits a security-relevant classifier's own detection data), so the full 5-axis fan-out ran (no safe-side skip).
• Correctness: no findings. The one stylistic redundancy noted (a same-shaped positive/negative assertion pair) was judged a pre-existing, deliberate convention in this file, not a defect.
• Blast-radius: no findings. No other script or test in the repository hardcodes _GOVERNANCE_FILENAMES's prior contents or length; no caller behavior changes unexpectedly.
• Reuse/simplification: 3 findings, CONFIRMED and fixed in commit 638982f0 -- merged two same-pattern tests into one, dropped a redundant not in assertion the positive assertion already implied, shortened a comment that restated CODEOWNERS almost verbatim.
• Convention: 1 finding, CONFIRMED and fixed in commit 638982f0 -- a new comment block's line-wrap width (80-89 chars) deviated from this file's own ~75-78 char convention; resolved as part of the same merge that shortened the comment.
• Security: no findings in this diff itself. Noted, out of this diff's scope, that .github/rulesets/main.json's existing required_approving_review_count: 0 setting is a pre-existing repository configuration choice this PR does not touch or introduce.
• Re-verification after the fix commit (638982f0): target test file 16/16 passed, full repo suite and local-preflight (47 gates) green; the merged/shortened diff was read again directly and confirmed to still satisfy every finding above with no new issue introduced.
• 0 confirmed findings outstanding as of 638982f0.

Merge gate: independent review

This PR is also subject to the independent-review-pending required
status check (see .github/workflows/independent-review-pending.yml /
.github/scripts/gitapex_gate_independent_review_pending.py). It stays
pending/failing until a ## Independent review verdict section naming
this PR's current head commit is recorded in this body --
drafting-a-pr-to-merge's own Step 8 records it once its independent
review completes. There is nothing for you to do here now: do not
pre-fill this section yourself, and do not remove this note.

Related Issue

Closes #1875

Execution log

• PlanApproved{run_id: 2026-09-06-governance-classifier-trusted-bots}
• StageDeviated{run_id: 2026-09-06-governance-classifier-trusted-bots, task_id: A, reason: "screening-a-low-trust-contribution check 4 (hook/script path modification) hard-flagged this task's own diff -- both changed files (gitapex_check_canonical_governance_paths.py, test_gitapex_check_canonical_governance_paths.py) live under skills/*/scripts/**, an unconditional-escalation category per that skill's own rules regardless of how benign the change reads. Content is purely additive and non-weakening (adds one _GOVERNANCE_FILENAMES entry + one regression test; strengthens rather than weakens the classifier), but executing-a-branch-plan Step 6/7 requires escalation on any screening flag with no retry.", action: escalate}
• Repository owner confirmed intentionality (in-session approval) to proceed with commit and push despite the check-4 flag.
• TaskCompleted{run_id: 2026-09-06-governance-classifier-trusted-bots, task_id: A, commit_sha: 36563c947fec282d96cb51980c5c506b7d7cd381}
• Step 8 refactor/simplify pass (branch-plan-task): reviewed both files' new lines against sibling conventions (comment style, test assertion pairing, placement, naming); found no genuine simplification; made no edits. Full suite + local-preflight re-confirmed green.
• Step 8 independent adversarial review (review-persona): CONFIRMED one completeness-gap finding -- .github/rulesets/main.json shares the same CODEOWNERS-documented risk class as trusted-bots.yml (issue feat(ci): exempt trusted bots from independent-review-pending's human-verdict requirement #1858's second trust anchor) but was left uncovered by this classifier. No other defect found (path-normalization, _HOOK_SCRIPT_PREFIXES interaction, and rationale-accuracy checks all cleared).
• Repository owner approved widening this PR's scope to also add .github/rulesets/main.json (issue fix(governance-paths): recognize .github/trusted-bots.yml as governance-sensitive #1875's own "Scope amendment" section records this).
• TaskCompleted{run_id: 2026-09-06-governance-classifier-trusted-bots, task_id: A2, commit_sha: 1ebd9207}: added .github/rulesets/main.json to _GOVERNANCE_FILENAMES plus regression test. Commit-message provenance scan clean.
• origin/main drift check (issue Local pre-push mirror gap, root-caused: session-start.sh never installs the pre-push shim (consolidates #1336/#1361/#1362) #1387) ran before pushing 1ebd9207: branch was 4 commits behind origin/main (unrelated review-persona.md call-site fix, PR fix(review-persona): add merge-retrospective Step 4b to Sanctioned call sites #1880); merged cleanly (no conflict); re-ran the target test file before pushing.
• CI's skill-audit-disclosure check flagged this PR's own ## Skill audit evidence section for two gaps: the checker-script-adversarial-review line's verdict token was not on its own recognized shape (a trailing period defeated the disclosure-line parser), and a required defeat-test-disclosure line was missing entirely. Fixed by adding two defeat tests (commit 6bd861b0) and rewriting this section in the parser's exact recognized shape.
• TaskCompleted{run_id: 2026-09-06-governance-classifier-trusted-bots, task_id: A3, commit_sha: 6bd861b0}: added the two defeat tests above. origin/main drift check re-run before pushing: 0 commits behind. Target test file, full repo suite, and local-preflight all pass.
• drafting-a-pr-to-merge Step 8 (inner layer, reviewing-an-artifact, low effort) ran against commit 1ebd9207's diff via a 5-axis fan-out (review-persona dispatch per axis). CONFIRMED 3 reuse/simplification findings and 1 convention finding (see Independent review verdict section above); 0 correctness/blast-radius/security findings.
• TaskCompleted{run_id: 2026-09-06-governance-classifier-trusted-bots, task_id: A4, commit_sha: 638982f0}: fixed all 4 CONFIRMED findings (merged two tests, dropped a redundant assertion, shortened two comments). origin/main drift check re-run before pushing: 0 commits behind. Target test file (16/16), full repo suite, and local-preflight (47 gates) all pass. Outer-layer Copilot review requested; repository owner directed proceeding without awaiting its response.

Note: the task's first commit attempt carried a Claude-Session trailer, which gitapex_check_task_commit_provenance.py hard-flagged (CONTRIBUTING.md's PR-body trailer exception does not extend to commit messages). Amended locally to drop that trailer before pushing; every pushed commit scans clean.

The ruleset-scan CI failure on an earlier head (b20c9bc) was confirmed by the repository owner to be an unrelated infrastructure issue (an astral-sh/setup-uv manifest fetch failure), not caused by this PR's diff.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: 022627a5-f2fd-4c96-bbbc-877f7ebf1cc2

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.56%. Comparing base (e6722d9) to head (638982f).
⚠️ Report is 7 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1881   +/-   ##
=======================================
  Coverage   99.56%   99.56%           
=======================================
  Files         168      168           
  Lines       27975    27975           
  Branches     3390     3390           
=======================================
  Hits        27852    27852           
  Misses        123      123           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[tvna]

Escalation: per-task screening hard flag (check 4)

executing-a-branch-plan Step 6's per-task screening dispatched this task's own diff (adding .github/trusted-bots.yml to _GOVERNANCE_FILENAMES in gitapex_check_canonical_governance_paths.py, plus a regression test) to screening-a-low-trust-contribution checks 2-8 via a review-persona subagent.

What was tried: the implementation was written and locally verified (pytest on the target test file: 14/14 passed; full repo suite: 8738/8738 passed; local-preflight: 47/47 gates passed; manual CLI check confirmed .github/trusted-bots.yml now classifies as governance). Before committing, the diff was screened per Step 6's mandatory per-task screening.

What the screening found: checks 2, 3, 5 (N/A -- no dependency manifest touched), 6, 7, 8 all cleared. Check 4 (hook/script changes) hard-flagged this diff, because both changed files live under skills/*/scripts/** -- one of this repository's own illustrative hook/script path categories, escalated unconditionally regardless of how the change reads. The reviewer noted the change is purely additive and strengthens (never weakens) the classifier, but per screening-a-low-trust-contribution's own Stop boundaries, a hard flag is never self-cleared on that basis.

Why this is escalated rather than fixed and committed: executing-a-branch-plan Step 7 states "a screening flag ... -> escalate ... these are not proof-method failures and get no retry at all," and Step 6 states "a flagged diff never proceeds to commit." This flag appears structurally unavoidable for this specific issue (#1875) -- the very target file this fix touches (the governance-path classifier itself) sits under the same skills/*/scripts/** prefix that check 4 treats as sensitive -- but the skill's own rules require a human decision before proceeding rather than this session silently treating that as a self-evident exception.

Current state: the diff is written and verified in the working tree, but not committed or pushed to this branch. Releasing the branch-plan-executing label now, per Step 7's ordering rule (comment first, then release).

Requesting: confirmation to proceed with committing and pushing this diff as-is (task scope matches the Branch Plan/ACM in issue #1875 and docs/gitapex/plans/2026-09-06-claude-governance-classifier-trusted-bots-dknenw.md), or a different instruction.

---

Generated by Claude Code