Claude Skills — Coleção de Skills Personalizados
bkizzy/claude-skills
Skills prontos para estender Claude Code com automação e relatórios
Ver no GitHubConjunto de skills customizados para Claude Code que estendem funcionalidades via comandos de barra (`/`). Inclui automação de dashboards (AAARRR, Phoenix), geração de documentos HTML, avaliação de RFPs e captura de screenshots. Cada skill é independente, versionado e documentado para facilitar instalação via symlink.
Instalação
npx convez-ai bkizzy/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
- Gerar relatórios AAARRR automatizados do App Store Connect e Play Console
- Avaliar RFPs com rubric estruturado e render HTML com análise de risco
- Capturar e organizar screenshots macOS em galeriaHTML auto-atualizável
- Criar documentos HTML self-contained como formato padrão em Claude
- Integrar observability com Phoenix LLM e gerar relatórios de erros e latência
Veredito

Repositório bem estruturado de skills production-ready com documentação clara e instalação simples via symlink. Foco em automação de workflows reais (métricas, avaliação de propostas, observability), embora exija familiaridade com Claude Code e arranjo local de pastas (~/.claude/skills). Contribuição sólida ao ecossistema.
README
A tradução para o português deste README ainda está na fila.
Claude skills
Custom Claude Code skills I've built. Each lives in its own folder with a SKILL.md (what Claude reads) and a README.md (what humans read).
Skills
| Skill | What it does |
|---|---|
aaarrr/ | 🏴☠️ App-store metrics, for pirates — daily AAARRR report. Claude drives the Claude-in-Chrome extension to read App Store Connect + Play Console dashboards you're already signed into. No API keys. |
html-doc/ | Makes HTML the default document format — a single self-contained file with a click-to-copy button. Falls back to .md/.docx/.pdf/.xlsx/.pptx only when requested. |
phoenix-errors/ | 🔭 Arize Phoenix error reports — Phoenix is an open-source LLM-observability/tracing platform; this skill walks you through connecting it to Claude as an MCP, then pulls ERROR spans across projects (multi-select) and renders an Excel + HTML report with a diagnostics view (error rate, latency, token usage, redelivery-loop detection). See the sample report. |
rfp-evaluator/ | Evaluates an RFP — reads the document (incl. charts/plans), scores it with a rubric + red-flag gates, and renders a role-tailored bid/no-bid HTML report. Document-grounded, no web search. |
screenshotr/ | 📸 Latest screenshots into the session — auto-detects your macOS screenshot folder and pulls the newest N. A cheap Haiku subagent reads, downscales, and describes them — so the main thread only learns what each image is and where it lives. Optionally --opens them in Preview or renders an auto-refreshing --html gallery in Claude's preview pane. |
Install
Skills are discovered from ~/.claude/skills/. To install one of these in your own Claude Code:
# Clone the repo somewhere (anywhere — pick your skills workspace)
git clone <repo-url> ~/code/claude-skills
# Symlink each skill you want into ~/.claude/skills/
ln -s ~/code/claude-skills/aaarrr ~/.claude/skills/aaarrr
Restart Claude Code (or /clear). Then invoke with /<skill-name> — e.g. /aaarrr.
Authoring notes
SKILL.mdis the instructions Claude reads at runtime. Frontmatter must includename(matches folder name = slash command) anddescription(drives auto-triggering).scripts/,references/,assets/are loaded on demand by the SKILL.md.- Per-skill
README.mddocuments the human-facing contract: what it does, how to use it, known gaps, where the rot points are.