agentsclimarketplace
Sign in

Verify

Skill TarasTsavolyk/claude-code-frontend/.claude/skills/verify

Run the quality gate and confirm a change actually meets its goal before commit — lint, tests, typecheck, plus a rules sanity pass. Use to close out any task ("make sure it works") rather than stopping at "should work".From its SKILL.md

Install
npx -y skills add TarasTsavolyk/claude-code-frontend --skill verify

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.4 KB, as published. Nobody here has run it

Verify

Turn "should work" into "verified". The quality gate is the floor (CLAUDE.md); don't stop until it passes (see CLAUDE.md → Working principles).

1. Run the gate

  • <pm> run lint && <pm> run test — and <pm> run typecheck in TS projects. All must pass; a failing gate blocks the commit.
  • e2e specs touched, or the change hits a flow only e2e covers → also <pm> run test:e2e (excluded from the default gate for speed, not because it's optional).
  • If a script name differs in this repo, use the actual one (package.json).

2. Confirm the goal, not just green

  • Restate the task as a concrete, checkable outcome and confirm the change meets it — new behavior is exercised by a test; a bug fix has a regression test that failed before and passes now (see testing.md).
  • Check the diff for scope creep: only what the task required changed (see CLAUDE.md → Working principles, surgical changes). No stray console.log, debug code, or commented-out blocks; no leftover .only/.skip in tests — a local .only silently shrinks the suite (Vitest allows it outside CI), so the gate goes green on one test.

3. See it, don't just infer it

  • UI changed and browser tooling is available (Playwright / Chrome DevTools MCP, or the running dev server)? Open the changed view and look: it renders, the async states actually appear, no console errors. Screenshot UI changes for the report.
  • No browser tooling? Note "verified by tests only" in the report instead of silently skipping — it's a different claim than "seen working".

4. Spot-check the relevant rules

  • New behavior: loading/error/empty/success states rendered, errors not swallowed (data-fetching.md, error-handling.md).
  • UI: keyboard-operable + labelled (accessibility.md); tokens not magic values (styling.md); no obvious XSS sink (security.md).
  • For a deeper pass, spawn the matching auditor agents (ui-reviewer, accessibility-auditor, performance-auditor, security-scanner) per workflow.md.

5. Report

  • State plainly what passed and what (if anything) is still failing or skipped — show the failing output, don't paper over it. Only call it done when the gate is green and the goal is met.

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.