qa
Skill0 starsChanged 25 days agoMITView on GitHub
Verify or reproduce behavior for real: run the project's tests, linters and type checks, check each acceptance criterion pass/fail, and boot the app locally (backend and/or frontend) to drive HTTP requests and a headless Playwright browser. Use on "test this", "verify", "run the QA plan", "test locally", "reproduce this bug locally", "does it work?", "/qa".
In this repository18 other files it gives its agents.
- address-commentsSkill
- demoSkill
- deploySkill
- documentSkill
- gather-contextSkill
- implementSkill
- investigateSkill
- observeSkill
- pr-follow-upSkill
- project-profileSkill
- pr-reviewSkill
- qa-planSkill
- resolve-conflictsSkill
- specSkill
- sprintSkill
- standupSkill
- todoSkill
- triageSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 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.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Follow ../../shared/policy.md.
- Very specific: line 24
Run the profile's commands.test scoped to the touched area first, then the wider suite if fast enough, then commands.lint / type checks / …
- Distinctive: line 33
Services: follow the user's and project's CLAUDE.md for local services (check before starting, stop what you started).
- Distinctive: line 60
Keep only the final useful repro script, in the work folder.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.
Files
QA
Follow ../../shared/policy.md. Run dev-skills:project-profile if not loaded this session.
Project-agnostic: discover how the project works from the profile and its docs; don't assume a stack.
Modes: verify (default when the branch has changes): prove the work meets its criteria.
repro: reproduce a reported scenario (used by investigate).
1. What to check
In order of preference: a QA plan given or found in the work folder; acceptance criteria in
context.md; the conversation; the branch diff against the base branch.
2. Automated checks
Run the profile's commands.test scoped to the touched area first, then the wider suite if fast
enough, then commands.lint / type checks / pre-commit hooks. Fix failures your change caused (when
called from a workflow that owns the change); report others as pre-existing with the decisive line.
3. Real run (when a criterion needs it)
Skip when automated tests already prove every criterion; say so.
- Layout: the project may be one app or frontend + backend as sibling/child folders. Signals:
package.jsonwith dev script,vite.config.*,next.config.*(frontend);Gemfile,manage.py,pyproject.toml,go.mod(backend); compose files,Procfile,Makefile. - Services: follow the user's and project's
CLAUDE.mdfor local services (check before starting, stop what you started). - Env: in a worktree, copy the main checkout's env files listed in the profile (
env_files). - Boot: documented commands first. Backends before frontends; point the frontend at the local backend. Run servers in the background with logs to a file in the work folder.
- API: curl or a short script; assert status and response shape.
- UI: Playwright headless against the local URL with a seeded/known login (seed files or test credentials from the project's fixtures, never real credentials).
repro: confirm the symptom first;verify: confirm the expected behavior (and, for a bug fix, that the symptom is gone).
4. Gotchas
- Port in use:
lsof -nP -iTCP:<port> -sTCP:LISTEN; don't kill unrelated processes, use another port. - macOS AirPlay Receiver holds port 5000.
- Stale servers from earlier runs: reuse or stop them.
- Per-tenant subdomains:
lvh.me/127.0.0.1.nip.ioresolve to localhost. - Boots but 500s: usually missing env/config.
New project-specific gotchas you hit go under
## Mine→### qain the profile, never in the repo.
Output
Checks: <command> → pass | fail (<decisive line>)
Criteria:
- [x] AC1 <text> — <evidence: test name, request/response, screenshot path>
- [ ] AC2 <text> — <what failed>
Real run: <done | skipped (why)> — tiers: <backend/frontend>Keep only the final useful repro script, in the work folder. Tear down servers you started.
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.

