Website: wst2002.github.io/claude-code-skills
Ready-to-use Claude Code skills and plugins that make AI coding agents faster, more reliable, and honest about their work.
Skills — all of them, or pick one:
npx skills add WST2002/claude-code-skills
npx skills add WST2002/claude-code-skills --skill security-auditOr copy any folder from skills/ into ~/.claude/skills/
(Windows: %USERPROFILE%\.claude\skills\).
Plugins — inside Claude Code:
/plugin marketplace add WST2002/claude-code-skills
/plugin install issue-to-pr@claude-code-skills
/plugin install proof-of-done@claude-code-skills
| Plugin | What it does |
|---|---|
| issue-to-pr | /fix-issue 123 — from a GitHub issue, Jira ticket or description to a tested fix and a draft pull request, in one command |
| proof-of-done | Blocks the agent from finishing after code changes until tests, build or lint have passed after the last edit, and catches tests skipped, weakened or deleted to get there |
proof-of-done in a real Claude Code session: Claude skipped the tests, the plugin made it run them, and they found a bug it had missed.
| Skill | Use it to |
|---|---|
| code-review | Review a PR, branch or diff for real defects, each with a concrete failing scenario |
| bug-debugger | Find the root cause of a bug, fix it, and prove the fix with a regression test |
| test-writer | Write tests that fail when behavior breaks, in your project's framework and style |
| refactor-planner | Refactor in small, safe, test-backed steps without changing behavior |
| proof-of-done-lite | Make the agent show test and build results before saying "done" |
| Skill | Use it to |
|---|---|
| test-case-generator | Turn a requirement or user story into manual test cases: positive, negative, boundary, permission and state |
| test-plan-writer | Write a risk-based test plan with scope, environments and measurable entry/exit criteria |
| bug-report-writer | Write bug reports developers can reproduce on the first try |
| playwright-e2e-tests | Write reliable Playwright end-to-end tests with user-facing locators and no sleeps |
| api-test-writer | Write API tests for status codes, schemas, validation, auth and side effects |
| flaky-test-fixer | Reproduce and fix tests that pass and fail without code changes |
| Skill | Use it to |
|---|---|
| codebase-onboarding | Understand an unfamiliar codebase fast: how to run it, the map, one real request flow |
| dependency-upgrade | Upgrade packages and frameworks in safe steps, following migration guides |
| changelog-writer | Write release notes and pick the next version from git history |
| Skill | Use it to |
|---|---|
| security-audit | Find exploitable vulnerabilities in any codebase, with file:line evidence, attack scenario and fix |
| web-performance-audit | Find why a site is slow, tied to LCP, INP, CLS and bundle size, with expected gains |
| accessibility-audit | Check keyboard, screen reader, contrast and ARIA issues mapped to WCAG 2.2 AA |
| seo-audit | Make pages crawlable and indexable: meta tags, canonical, sitemap, structured data |
For designers too: these skills also work in the Claude app, with a screenshot, a Figma frame or your research notes. No terminal needed:
- Download the skill's ZIP from its page on the website.
- In Claude, turn on Code execution (Settings → Capabilities).
- Open Customize → Skills, click +, choose Upload a skill and pick the ZIP.
design-review turns a screenshot into a shareable report: numbered pins on the screen, exact contrast ratios measured from the pixels, and a concrete fix for each finding. Open the example report.
| Skill | Use it to |
|---|---|
| design-review | Get senior-designer feedback on a screen: usability, hierarchy, accessibility and missing states, each with a concrete fix, as a report with numbered pins on the screenshot |
| ux-research-synthesis | Turn interview notes and transcripts into findings with quotes, participant counts and "how might we" opportunities |
| design-handoff-spec | Write the developer handoff: every state, edge case, breakpoint and interaction, with open questions instead of guesses |
| ux-writing | Write or review buttons, error messages, empty states and other UI text: clear, short and consistent |
| figma-design-compare | Check a website, web app or mobile screen against its Figma frame: screenshots plus exact font, color, spacing and size values |
| Skill | Use it to |
|---|---|
| database-schema-design | Design a schema from scratch or review one for integrity, indexes and safe migrations |
| api-design-review | Design or review REST APIs: naming, status codes, errors, pagination, breaking changes |
| architecture-diagram | Draw system, sequence, ERD or deployment diagrams in Mermaid, from the real code |
| Skill | Use it to |
|---|---|
| dockerize-app | Write a small, secure Dockerfile and docker-compose for Node, Python, Java or Go, and verify it runs |
| github-actions-ci | Set up or fix GitHub Actions CI with caching, services and least-privilege permissions |
Every skill uses the same approach: a clear method, a fixed output format, and a verification step so results are specific and low on false positives.
Tested: skills are benchmarked with and without the skill on sample projects with seeded bugs and answer keys — see benchmarks.
- Skills are instructions. Claude loads them when your request matches.
- Plugins can enforce behaviour with hooks, for example stopping the agent until checks pass.
See CONTRIBUTING.md. Ideas for new skills are welcome as issues.
MIT © Wageesha Tripathi


