piv-validate

name	piv-validate
description	Runs this project's full validation suite — tests, type checks, and linting across every part of the stack — then reports overall health. Use before committing, before opening a PR, or after finishing a chunk of work to confirm zero regressions.

Validate

Run every check this project has and report a single PASS/FAIL verdict.

⚠️ Make this yours first

This skill is a template. The command list below is a placeholder — replace it with the commands your project actually uses. That is the entire point of a custom checker: it is the one skill that cannot be generic, because it wraps your stack's CLIs.

Find your real commands in package.json scripts, pyproject.toml, Makefile, justfile, docker-compose.yml, your CI workflow, or the README — then delete these and paste yours in.

Two things to get right, because they are the usual reason a checker silently lies:

1. Working directory. Many tools only discover their config from the current directory. If your config lives in backend/pyproject.toml, the command is cd backend && uv run pytest, not uv run pytest from the repo root.
2. Cross-platform. Avoid POSIX-only idioms if anyone on the team is on Windows — lsof, python3, and background-then-kill are the common offenders.

Run the checks in order. Keep going after a failure so the report covers everything, and capture the output of any command that fails.

1. Tests

# REPLACE: your test command, run from the right directory
<your test command>

Expected: all tests pass.

2. Type check

# REPLACE: e.g. mypy, tsc --noEmit, go vet
<your type-check command>

Expected: no type errors.

3. Lint / format check

# REPLACE: e.g. ruff check, biome check, eslint
<your lint command>

Expected: clean.

4. Repeat per surface

A full-stack project has more than one of each. Add a section per surface — backend tests, backend types, backend lint, frontend tests, frontend types, frontend lint — so a single command covers the whole repo.

5. Optional — live smoke test

Only when the change touches routing, middleware, or startup; skip it when your test suite already exercises the app in-process.

# REPLACE: start your app, hit one endpoint, confirm the status code, stop it
<your run command>

Prefer starting the server in a second shell over backgrounding and killing it from inside this skill — the background-and-kill idiom is not portable.

6. Summary report

Report each check with a ✅ or ❌, then an overall verdict:

• One line per check
• Overall: PASS or FAIL

For every ❌, include the failing command and the relevant output. Do not fix anything here — this skill reports; fixing is a separate step.

Notes

• Keep this skill fast. It runs before every commit; if a step gets slow, that is a signal to fix the slow step, not to drop it from the checker.
• A checker that cannot fail is worthless. Once your commands are wired in, break something on purpose and confirm this skill reports ❌.