convez.ai
SkillDevOps & InfraestruturaAvançado
pathiec92

Claude Skills para Engenheiros Sênior

pathiec92/claude-skills

Playbooks de principal engineer para arquitetura, code review e ML systems

Ver no GitHub
InicialPluginScore · de 0 a 100

Coleção curada de skills reutilizáveis para Claude Code focada em arquitetura cloud-native, MLOps e aplicações LLM/AI. Cada skill é um framework executável que Claude invoca automaticamente quando seu request corresponde ao gatilho, eliminando a necessidade de colar checklists manualmente.

Instalação

$npx convez-ai pathiec92/claude-skills

1 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

Estrelas
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Revisar arquiteturas de sistemas distribuídos e cloud-native
  • Code review em nível de principal engineer com foco em observabilidade
  • Design de sistemas de ML/RAG/LLM em produção
  • Documentar decisões arquiteturais com ADRs
  • Análise pós-incidente e postmortems

Veredito

Excelente para engenheiros sênior que precisam de frameworks executáveis em vez de documentação. A seleção de skills é bem pensada com gatilhos explícitos, mas requer adoção em Claude Code (ferramenta específica) — portabilidade limitada a outros ambientes.

README

A tradução para o português deste README ainda está na fila.

claude-skills

A curated set of Claude Code skills for senior engineers and architects working on cloud-native platforms, distributed systems, MLOps, and LLM/AI applications.

Each skill is a focused, reusable framework that Claude invokes automatically when your request matches it — so the right principal-engineer playbook shows up at the right moment, without you having to paste a checklist every time.


What's a skill?

A skill is a SKILL.md file with YAML frontmatter (name + description) and a markdown body. Claude reads the description to decide when to apply the skill, and the body for how. These skills are designed with explicit trigger boundaries (each says when to use it and when to defer to a sibling), so selection stays unambiguous as the catalog grows.


Skill catalog

SkillWhat it doesTriggers when you…
architecture-reviewSystem/infra design review across reliability, scale, security, ops, costask "review this architecture", discuss SPOFs, DR, 10x/100x scaling
principal-code-reviewSource/diff review at a principal bar (correctness → observability)ask to review code, a PR, or hunt for bugs
gcp-architectureGCP service selection, cost, and IAM with a managed-first biasdesign on GCP, mention Cloud Run/GKE/Pub/Sub/BigQuery
ml-system-designEnd-to-end ML systems: data → training → serving → MLOpsdesign a recommender/ranking system, review an ML pipeline
rag-system-designProduction RAG: chunking, retrieval, reranking, grounding, evalbuild/debug RAG, semantic search, or Q&A-over-docs
llm-app-architectureProduction LLM apps: tools, agents, guardrails, eval, costbuild an LLM feature, agent, or assistant
mlops-pipeline-reviewData/training/registry/CI-CD-CT, deployment, drift, rollbackproductionize a model or review an ML pipeline
observability-reviewMetrics, logs, traces, SLIs/SLOs, alertingadd instrumentation, define SLOs, cut alert noise
incident-postmortemBlameless RCA: timeline, contributing factors, action itemswrite a postmortem after an outage
adr-writerArchitecture Decision Records with options and trade-offsdocument a significant technical decision
leadership-lensOrg/team/governance framing of technical decisionsweigh platform vs point solution, standardization, tech-debt
interview-modeActs as a Principal interviewer; withholds answers, scores youpractice system-design / ML-system-design interviews
learning-coachTeaches at architect depth: why, trade-offs, failure modeswant a deep dive, not a tutorial

Install

Skills live in ~/.claude/skills/. Clone this repo and run the installer for your platform, or copy the folders manually.

macOS / Linux:

git clone https://github.com/pathiec92/claude-skills.git
cd claude-skills
./install.sh

Windows (PowerShell):

git clone https://github.com/pathiec92/claude-skills.git
cd claude-skills
./install.ps1

The installer copies each skill into ~/.claude/skills/. Re-run it after pulling updates. Restart Claude Code (or start a new session) to pick up new skills.

Use just one or two

You don't need all of them — each skill is a self-contained folder. Copy only the ones you want:

cp -r skills/rag-system-design ~/.claude/skills/

Design principles

These skills follow a few rules that make automatic selection reliable:

  • Single responsibility — one job per skill; no kitchen-sink skills.
  • Explicit boundaries — every description states when not to use it and which sibling to use instead (e.g. RAG → rag-system-design, not ml-system-design).
  • Action-oriented bodies — checklists and frameworks, not prose, so Claude produces structured output.
  • Always-on vs on-demand — these are on-demand frameworks; keep persistent preferences in your CLAUDE.md, not here.

About

Maintained by Pampapathi K — AI Platform / Cloud Architect focused on GCP, MLOps, and large-scale distributed systems. These skills distill the review frameworks and design checklists I use day to day.

Contributions and forks welcome.

License

MIT — use, modify, and share freely.

Ver o original no GitHub

Bom para

Relacionados