agentleFS
hushh-labshushh-labs/hushh-research/.agents/skills/pr-governance-review/SKILL.md

pr-governance-review

Skill40 starsChanged 5 days agoApache-2.0View on GitHub

Use when reviewing an incoming pull request for north-star alignment, trust-boundary regressions, malicious or low-signal degradation, stale-vs-current CI interpretation, and true merge readiness beyond a green gate. Also owns PR train planning, merge queue cohorts, maintainer patch-then-merge, contributor harvest decisions, and backlog review waves.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Avoid overlap with repo-operations, quality-contracts, backend-runtime-governance, frontend-architecture, and security-audit.

    line 12
  2. Distinctive:

    Run the delegation router at intake.

    line 67
  3. Very specific:

    Run the PR checklist or hybrid live report and treat contract_set, duplicate_group, public_comment_policy, lane, and live_report_action as …

    line 68
  4. Very specific:

    Check references/anti-rationalization.md at intake.

    line 71
  5. Very specific:

    Use comment-and-report-contract.md for every GitHub write.

    line 73
  6. Distinctive:

    Keep branch switching, commits, GitHub writes, approvals, merges, deploys, report refreshes, and final decisions in the parent session.

    line 74
  7. Very specific:

    Test placement or verification policy routes to quality-contracts.

    line 80

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

PR Governance Review Skill

Purpose and Trigger

  • Primary scope: pr-governance-review-intake
  • Trigger on PR review, merge readiness, maintainer patching, backlog train planning, PR governance reports, or contributor-facing PR decisions.
  • Avoid overlap with repo-operations, quality-contracts, backend-runtime-governance, frontend-architecture, and security-audit.

Coverage and Ownership

  • Role: owner
  • Owner family: pr-governance-review

Owned repo surfaces:

  1. .agents/skills/pr-governance-review

Non-owned surfaces:

  1. repo-context
  2. repo-operations
  3. quality-contracts
  4. backend-runtime-governance
  5. frontend-architecture
  6. security-audit

Do Use

  1. Current-head PR review beyond green CI.
  2. Async PR-train planning, queue cohorts, collision groups, patch trains, and decision waves.
  3. Maintainer patch/harvest decisions, contributor acknowledgement, and lane-specific GitHub comment posture.
  4. Review for north-star drift, trust-boundary regression, stale CI, duplicate architecture, or unreachable helpers.

Do Not Use

  1. Broad repo orientation before a PR or owner surface is known.
  2. CI root-cause repair after a check failure is already classified.
  3. Domain implementation work after a PR has been handed to an owner skill.

Read First

  1. docs/operations/releases/ci.md
  2. docs/development/quality/pr-impact-checklist.md
  3. .agents/skills/repo-operations/SKILL.md
  4. .agents/skills/pr-governance-review/references/review-axes.md
  5. .agents/skills/pr-governance-review/references/operator-batch-output-contract.md
  6. .agents/skills/pr-governance-review/references/blocker-gates.md
  7. .agents/skills/pr-governance-review/references/comment-and-report-contract.md
  8. .agents/skills/pr-governance-review/references/pr-train-review-sop.md

Workflow

Review against the author's intended protected destination: Dev for integration, UAT for a selected candidate, and Main for production-baseline or operations work. Ordinary contributors use Dev intake; governed maintainers may propose PRs directly into main by actor identity under the separate Main checks. UAT can qualify directly without prior Dev deployment. Existing train campaigns retain their declared purpose; do not retarget ordinary PRs to a train branch. Required current checks, exact combined-result validation, applicable team review and Admin SOP apply independently. A merge never deploys an application.

  1. Lock the current PR head SHA, current CI Status Gate, mergeability, draft state, and review state before judging.
  2. Run the delegation router at intake; use real read-only subagent evidence lanes for non-trivial, high-risk, or multi-PR work.
  3. Run the PR checklist or hybrid live report and treat contract_set, duplicate_group, public_comment_policy, lane, and live_report_action as decision records.
  4. Use the async PR governance subagent train method as the default for more than one PR: first lock the operator-approved surface scope, identify only trains inside that scope or its hard dependencies, run scoped non-touching trains in parallel through subagent evidence lanes, sequence touching PRs oldest-first inside each train, queue independent green PRs together, run disjoint patch trains, and run decision waves asynchronously. Unrelated green-clean PRs stay in out_of_scope_candidates until a separate operator checkpoint approves a broader sweep.
  5. Exclude PRs with failing/missing/stale required checks or failing auxiliary checks from executable trains unless the task is CI repair.
  6. Check references/anti-rationalization.md at intake; then apply blocker gates before merge: north-star drift, duplicate architecture, trust-boundary regression, caller/proxy/backend mismatch, unreachable helpers, stacked diff, proof gaps, and local dirty-file overlap.
  7. Prefer direct contributor PR merge, then maintainer_patch_then_merge, then maintainer harvest. Harvest is allowed only when the source PR should not be the merge vehicle and the plan names accepted value, canonical attach point, write set, dropped/deferred pieces, proof, source PR close-or-hold state, and co-author attribution when code/tests are materially reused.
  8. Use comment-and-report-contract.md for every GitHub write; edit existing maintainer records first and post one post-merge closeout after smoke.
  9. Keep branch switching, commits, GitHub writes, approvals, merges, deploys, report refreshes, and final decisions in the parent session.
  10. Preserve the current operating-history lessons in evolution-history.md; do not append incident history to this compact kernel.

Handoff Rules

  1. CI, queue, deploy, or branch-protection authority routes to repo-operations.
  2. Test placement or verification policy routes to quality-contracts.
  3. Backend runtime boundaries route to backend-runtime-governance.
  4. Frontend caller/route contracts route to frontend-architecture.
  5. IAM, consent, vault, PKM, or sensitive data boundaries route to security-audit.

Required Checks

python3 -m py_compile .agents/skills/pr-governance-review/scripts/pr_review_checklist.py .agents/skills/pr-governance-review/scripts/test_pr_review_checklist.py .agents/skills/pr-governance-review/scripts/build_runtime_schematics.py .agents/skills/pr-governance-review/scripts/contributor_impact_report.py
python3 .agents/skills/pr-governance-review/scripts/test_pr_review_checklist.py
python3 .agents/skills/agent-orchestration-governance/scripts/agent_router_smoke.py
python3 .agents/skills/pr-governance-review/scripts/build_runtime_schematics.py --text
python3 .agents/skills/pr-governance-review/scripts/pr_review_checklist.py --repo hushh-labs/hushh-research --live-report --scan-mode hybrid --selection-order oldest --limit 100 --candidate-limit 40 --text --output tmp/pr-governance-live-report.md
python3 .agents/skills/pr-governance-review/scripts/test_contributor_impact_report.py
./bin/hushh agents audit --text
./bin/hushh docs verify

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.