Claude Skills - QA, First Principles & Otimização PostgreSQL
yuribodo/claude-skills
Skills de QA, arquitetura e otimização PostgreSQL para Claude Code
Ver no GitHubColeção de skills para Claude Code que automatizam testes de QA end-to-end com GIFs, análise de problemas do zero, e otimização de queries PostgreSQL. Ideais para validar trabalho pronto antes de enviar, resolver bugs estruturais e diagnosticar queries lentas.
Instalação
npx convez-ai yuribodo/claude-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Executar testes end-to-end antes de considerar tarefa concluída
- Diagnosticar e otimizar queries PostgreSQL lentas
- Decompor problemas arquiteturais e tomar decisões de tech stack
Veredito

Skill bem estruturada e pragmática: o `qa` resolve um pain point real (validação antes do ship) com automação inteligente, incluindo GIFs e feedback acionável. O `postgres-query-optimization` traz método científico para problema comum. Recomendado para times que querem elevar padrão de qualidade em Claude Code.
README
A tradução para o português deste README ainda está na fila.
claude-skills
A collection of personal Claude Code skills.
| Skill | What it does |
|---|---|
qa | Senior-QA gate before calling a task "done": checks the work against its requirements, runs unit tests, drives the live app in a browser (logging in if needed), and returns a verdict + a GIF walkthrough, screenshots, console/network errors, and actionable dev feedback. Never commits, pushes, or fixes code. |
first-principles | Decompose problems to fundamental truths and rebuild from scratch. Use for architecture/tech-stack decisions, conventional-fix-resistant bugs, performance past surface caching, or questioning cargo-cult / hype-driven choices. |
postgres-query-optimization | Diagnose and fix slow PostgreSQL queries with a measure-first method: find offenders via pg_stat_statements, read EXPLAIN (ANALYZE, BUFFERS) plans, fix the root cause (indexes, sargability, stale stats, bad plans), and verify. |
Install
One-liner (recommended)
Install a single skill straight into ~/.claude/skills/:
curl -fsSL https://raw.githubusercontent.com/yuribodo/claude-skills/main/install.sh | sh -s -- qa
Pick any of: qa, first-principles, postgres-query-optimization (space-separate
to install several at once).
Install all skills in this repo:
curl -fsSL https://raw.githubusercontent.com/yuribodo/claude-skills/main/install.sh | sh
From a clone
git clone https://github.com/yuribodo/claude-skills.git
cd claude-skills
./install.sh qa # or: ./install.sh (installs everything)
The installer copies each skill into ~/.claude/skills/<skill>/. Set
CLAUDE_SKILLS_DIR to install somewhere else. Restart Claude Code (or start a
new session) afterwards so it picks up the skill.
Skills
qa — Senior QA before "done"
Use it when you're about to call a task complete, or just say "QA this", "test it end to end before I ship", or "be my QA".
It dispatches an isolated QA subagent (so your main chat stays clean) that:
- Establishes the requirements — explicit acceptance criteria from the task +
git diff, plus the implicit ones a senior QA always checks (error/empty/loading states, auth boundaries, input validation, persistence-after-refresh, responsive, a11y basics). - Runs the unit/logic layer — auto-detects the test runner, runs it, flags changed code with no coverage.
- Launches the app — reusing your project's run convention; asks you once for credentials if login is needed (never invents or hardcodes them).
- Drives the live app in a browser — happy path and negative/edge cases, recording a GIF, capturing screenshots of key states, and collecting console + network errors.
- Returns a verdict —
PASS/PASS-WITH-ISSUES/FAIL, with each bug in severity order: what the user sees · how to reproduce · expected vs actual · why it matters · likely subsystem to look at.
Hard rules baked in: report only — never fixes code, and never commits,
pushes, or opens PRs. Artifacts land in ./qa-artifacts/<date>/ in the repo under
test (add that to your .gitignore).
Files:
qa/
├── SKILL.md # trigger + dispatch layer
└── references/
└── qa-protocol.md # the full senior-QA protocol the subagent follows
first-principles — think from fundamental truths
Use it for architecture and tech-stack decisions, bugs that resist conventional fixes, performance work past surface-level caching, or whenever a choice smells like cargo-cult / resume-driven / hype-driven ("should we use X?", "everyone uses X for this", "why is this so complex?"). It decomposes the problem to what's actually true and rebuilds the solution from there.
first-principles/
├── SKILL.md
├── references.md
└── worked-examples.md
Standalone repo: yuribodo/first-principles-skill.
postgres-query-optimization — fix slow Postgres queries
A measure-first methodology: find the real offenders with pg_stat_statements, read
EXPLAIN (ANALYZE, BUFFERS) plans, fix the root cause (missing/wrong index,
non-sargable predicate, stale stats, bad plan), and verify the improvement. Use it
when a query is slow, when reading an execution plan, designing indexes, or chasing
sequential scans, sort spills, bad row estimates, or N+1 patterns.
postgres-query-optimization/
├── SKILL.md
├── ANTIPATTERNS.md
└── REFERENCE.md
Standalone repo: yuribodo/postgres-query-optimization-skill.
License
MIT