pre-pr
Skill0 starsChanged 4 months agoMITView on GitHub
Run the full pre-PR gate — lint, types, tests, migration heads, docs, spec citations, follow-ups — so CI passes on first push instead of going red
- Commits and pushes
10 other projects use a version of this.
Adopted by
Most recently taken up by
- safe-global/safe-wallet-monorepotheir own version, first seen Nov 9, 2026
- salvo8519/fairmarktheir own version, first seen Nov 8, 2026
- OrionArchitekton/orion-skillstheir own version, first seen Nov 2, 2026
- TickLabVN/biopasstheir own version, first seen Nov 2, 2026
- JohnnyMcGee/claude-skillstheir own version, first seen Nov 1, 2026
- plastic-labs/honchotheir own version, first seen Oct 23, 2026
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 10
- mongodb/mongo-go-driver★ 8.5ksame name
- plastic-labs/honcho★ 8.1ksame name
- TickLabVN/biopass★ 602same name
- safe-global/safe-wallet-monorepo★ 598same name
- asynkron/asynkron-skills★ 32same name
- MichinaoShimizu/kiroku★ 2same name
- OrionArchitekton/orion-skills★ 2same name
- Docat0209/SoloForge★ 1same name
- salvo8519/fairmark★ 1same name
- JohnnyMcGee/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository29 other files it gives its agents.
- adopt-existing-projectSkill
- adrSkill
- branch-conflict-checkSkill
- ci-watchSkill
- compound-learningsSkill
- config-auditSkill
- cost-checkSkill
- coverage-ratchetSkill
- doc-syncSkill
- eval-harnessSkill
- flaky-triageSkill
- followup-trackingSkill
- incident-captureSkill
- kit-doctorSkill
- kit-updateSkill
- logging-initSkill
- migration-checkSkill
- new-project-bootstrapSkill
- nightly-auditSkill
- observability-checkSkill
- parallel-workSkill
- perf-budgetSkill
- pr-babysitterSkill
- prompt-regressionSkill
- releaseSkill
- session-handoffSkill
- sync-healthSkill
- using-the-kitSkill
- writing-kit-skillsSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. 0 are common, 7 are distinctive, and 6 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
7
Distinctive
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Run this before every gh pr create, from inside the worktree you did the work in.
- Very specific: line 22
Strictness-aware: Read gates.strictness once at the start (see docs/PROFILE.md).
- Distinctive: line 31
Create one todo per step, then work them in order.
- Very specific: line 39
If it fails, run uv run ruff check --fix src/ tests/ and re-run.
- Distinctive: line 53
If your change touches an integration surface, also run:
- Distinctive: line 68
If the change touches auth, secrets, user input validation, or external calls (outbound HTTP, email, calendar, Discord), dispatch the …
- Distinctive: line 81
If false, skip and mark N/A.
- Very specific: line 83
Run the coverage-ratchet skill to confirm the PR does not drop coverage below the stored baseline (.coverage-baseline).
- Very specific: line 90
Run the perf-budget skill to confirm no budgeted hot path or model call regressed past its p95 tolerance versus the baseline.
- Distinctive: line 116
Run the frontend build and typecheck so a broken UI doesn't slip through.
- Very specific: line 127
Run the secret-scan-diff.sh hook over the branch diff to catch keys/tokens pasted into ordinary source files (the edit-time guard only …
- Distinctive: line 143
Run the prompt-regression skill — re-run the affected task_type's eval fixtures and confirm no gated tier regressed.
- Distinctive: line 155
Run the branch-conflict-check skill to see whether another open PR touches the same files.
“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Pre-PR Gate
Run this before every gh pr create, from inside the worktree you did the
work in. Its job is to catch everything CI would catch, locally, so no PR arrives
red and blocks the queue for other agents. Every check mirrors a CI job — keep
the two in lockstep: when you add a CI job, add a step here.
Profile-driven. Commands and capability gates below read from .claude/kit.yaml.
Each step tagged # kit.yaml → <key> runs the command at that key if set; an
empty string in the profile means skip the step and mark it N/A in the output.
Strictness-aware. Read gates.strictness once at the start (see
docs/PROFILE.md). At prototype, the trend gates — coverage ratchet, perf
budget, prompt-regression pass-gates — report their numbers but do not block
the PR; note (advisory — prototype) on their output lines. At production,
step 5's security review runs on every diff, not only sensitive surfaces.
Secrets scan, migrations, and security findings block at every level.
Checklist
Create one todo per step, then work them in order. Stop and fix on any failure before moving on — do not report "ready" with a known-red step.
1. Lint
# kit.yaml → toolchain.lint
uv run ruff check src/ tests/If it fails, run uv run ruff check --fix src/ tests/ and re-run. The kit's
PostToolUse hook already auto-fixes on save, so this should usually be clean.
2. Types
# kit.yaml → toolchain.typecheck
uv run mypy src/3. Tests
# kit.yaml → toolchain.test
uv run pytest tests/unit/ -m "not slow and not llm" --tb=short -qIf your change touches an integration surface, also run:
# kit.yaml → toolchain.test_integration
uv run pytest tests/integration/ -m "not slow and not llm" -q4. Test coverage of the change
Dispatch the test-gap-analyzer agent to confirm new public functions and changed branches have covering tests. Have it report every gap it identifies, each with a confidence and severity rating — not only the gaps it judges important. The blocking-vs-follow-up split below does the triage; the agent should not pre-filter. A new untested code path is a blocking gap; happy-path-only coverage of a risky path is a follow-up.
5. Security review
If the change touches auth, secrets, user input validation, or external calls
(outbound HTTP, email, calendar, Discord), dispatch the security-reviewer
agent over the diff. It audits for credential leaks, injection paths, auth bypass,
and insecure token handling. Have it report every finding with a confidence
and severity rating rather than filtering to only what it judges important —
the accept/block decision happens in this step, not by the agent silently
withholding lower-severity findings. Any finding blocks the PR until addressed
or explicitly accepted with a documented reason. If none of those surfaces are
involved, mark N/A — except at production strictness, where the review runs
on every diff regardless of surface.
6. Coverage ratchet
Applies when capabilities.coverage.ratchet_enabled is true in .claude/kit.yaml.
If false, skip and mark N/A.
Run the coverage-ratchet skill to confirm the PR does not drop coverage below
the stored baseline (.coverage-baseline). A regression blocks the PR; an
accepted dip requires an updated baseline and a note in the follow-ups log.
7. Performance budget
Applies when capabilities.perf.enabled is true. If false, skip and mark N/A.
Run the perf-budget skill to confirm no budgeted hot path or model call
regressed past its p95 tolerance versus the baseline. A regression blocks the PR;
an accepted change requires an updated .perf-baseline and a follow-up note.
8. Migrations
Applies when capabilities.migrations.enabled is true. If false, skip and mark N/A.
# kit.yaml → capabilities.migrations.heads_cmd
uv run alembic heads # expect exactly ONE headThen run the migration-check skill — it catches destructive/irreversible ops (dropped columns, non-nullable-without-default, narrowing) before they reach the primary DB and any replica. More than one head means a merge migration is needed.
9. Build / typecheck any UI
Applies when capabilities.ui.enabled is true. If false, skip and mark N/A.
# kit.yaml → capabilities.ui.build_cmd
<ui-build-command># kit.yaml → capabilities.ui.typecheck_cmd
<ui-typecheck-command>Run the frontend build and typecheck so a broken UI doesn't slip through.
10. Build artifact
Applies when toolchain.build is non-empty. If empty, skip and mark N/A.
# kit.yaml → toolchain.build
<build-command>11. Secrets scan
Run the secret-scan-diff.sh hook over the branch diff to catch keys/tokens pasted into ordinary source files (the edit-time guard only covers credential files). Any hit blocks the PR until removed and rotated.
12. Docs sync
Dispatch the docs-updater agent (or run the doc-sync skill) to update any
docs affected by the change. Narrative docs, changelog, and the API reference
must not drift from the code in the same PR.
13. Spec + drift
Dispatch the spec-drift-checker agent. Every design change must cite the
spec section it implements, and any behavior that diverges from the spec must
either update the spec in this PR or be logged in the follow-ups file with a
reason. If capabilities.spec.file is empty in kit.yaml, mark N/A.
14. Prompt regression (if the change touches a prompt or model config)
Run the prompt-regression skill — re-run the affected task_type's eval fixtures and confirm no gated tier regressed. An accepted score tradeoff must be logged as a follow-up.
15. Follow-ups & learnings
Scan your diff for TODOs, deferred decisions, and accepted drift. Append them to
the follow-ups log (doc-sync skill / docs/followups.md). Close any follow-up
this PR resolves. If the work solved a non-obvious problem (a bug that took real
investigation, a surprising dependency constraint, a workaround), also run the
compound-learnings skill so the solution lands in docs/solutions/.
16. Branch-conflict check
Run the branch-conflict-check skill to see whether another open PR touches the same files. If so, coordinate merge order before opening this one.
17. Working tree
git status --porcelain # expect empty after commitOutput
Fill this in with real results — never check a box you didn't verify. Mark
conditional gates N/A when they don't apply:
## Pre-PR Gate
- [ ] Lint: <clean / N fixed>
- [ ] Types: <clean / errors>
- [ ] Tests: <XX passed>
- [ ] Integration tests: <XX passed / N/A>
- [ ] Test coverage: <no gaps / N new untested paths>
- [ ] Security review: <clean / N findings addressed / N/A>
- [ ] Coverage ratchet: <pass / dropped N% below baseline / N/A>
- [ ] Performance budget: <within budget / regressed <path> / N/A>
- [ ] Migrations: <1 head, safe / N/A>
- [ ] UI build + types: <clean / N/A>
- [ ] Build: <clean / N/A>
- [ ] Secrets scan: <clean / BLOCKED — leak at ...>
- [ ] Docs: <updated <files> / no docs affected>
- [ ] Spec: <cited §X.Y / §X.Y may need update — reason / N/A>
- [ ] Prompt regression: <no regression / N/A>
- [ ] Follow-ups: <none / added N / closed N>
- [ ] Learnings: <captured docs/solutions/<file> / none to capture>
- [ ] Branch conflicts: <none / overlaps PR #N on <files>>
- [ ] Working tree: <clean>
Ready for PR: YES / NO — <blocking issues>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
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.

