Claude Senior Engineer Skills
ramzi-bouzaiene/claude-senior-engineer-skills
30 skills que transformam Claude no seu engenheiro sênior de código
Ver no GitHubColeção de 30 skills estruturadas que codificam o pensamento de engenheiros sênior em instruções reutilizáveis. Cobre todo ciclo de vida do código: compreensão, debug, correção, refatoração, otimização, hardening e revisão. Cada skill inclui técnicas avançadas de raciocínio (protocolos de pensamento, auto-verificação, avaliação de confiança).
Instalação
npx convez-ai ramzi-bouzaiene/claude-senior-engineer-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
- Debug estruturado e rastreamento de problemas complexos
- Refatoração segura com preservação de comportamento
- Análise de fluxo de código e identificação de áreas de risco
- Otimização de performance com justificativa técnica
- Hardening e segurança de aplicações
- Code review arquitetural com raciocínio explícito
Veredito

Recurso excepcional para engenheiros que buscam aumentar produtividade e qualidade do código usando Claude como parceiro técnico de confiança. A integração de técnicas avançadas de raciocínio (evidência, hipótese, confiança) eleva o padrão de assistência IA muito acima do comum, tornando-a genuinamente útil em decisões técnicas reais.
README
A tradução para o português deste README ainda está na fila.
Claude Senior Engineer Skills
From Bug → Clean → Scalable → Production
A curated collection of 30 Claude-style skills that mirror how senior engineers actually think about code, not just writing it, but understanding, debugging, fixing, refactoring, optimizing, hardening, and reviewing it.
These skills are designed to be used with Claude Code or any Claude-powered AI workflow. Each skill is a precise, opinionated instruction set that transforms Claude into a domain-specific expert for a given engineering task.
🧠 The Philosophy
Most AI coding prompts are shallow. They say "fix this bug" or "refactor this function" without the depth a senior engineer brings to the table.
This repository encodes the mental models, constraints, and judgment of a seasoned engineer into reusable, composable skill files. The result: AI assistance that doesn't just produce code, it produces correct, safe, maintainable code backed by real engineering thinking.
🚀 What's New: Advanced AI Techniques
Every skill now includes advanced prompting techniques that dramatically improve Claude's reasoning quality:
| Technique | What It Does |
|---|---|
| 🧠 Reasoning Protocol | Forces explicit step-by-step thinking: Observe → Recognize Patterns → Form Hypothesis → Gather Evidence → Conclude |
| ✅ Self-Verification Checklist | Built-in quality gate: evidence grounded? assumptions stated? edge cases covered? |
| 🎯 Confidence Assessment | Tags findings as 🟢 High / 🟡 Medium / 🔴 Low confidence with required actions |
| ⚔️ Adversarial Self-Review | "What would a skeptical senior engineer challenge?" — addresses strongest objection |
Why This Matters
Traditional prompts produce answers. These skills produce auditable reasoning:
❌ Old: "The bug is in line 42"
✅ New: "Based on [evidence], I hypothesize [X]. This is supported by [Y] but
contradicted by [Z]. Confidence: Medium. Verify by [action]."
🔄 The Pipeline
| Stage | Folder | Skills | Purpose |
|---|---|---|---|
| 🔍 Understand | 01_understand/ | 4 | Map, analyze, and explain existing code |
| 🐛 Debug | 02_debug/ | 4 | Reproduce, trace, and diagnose issues |
| 🔧 Fix | 03_fix/ | 4 | Apply targeted, safe, minimal patches |
| ♻️ Refactor | 04_refactor/ | 4 | Improve structure without changing behavior |
| ⚡ Optimize | 05_optimize/ | 4 | Eliminate bottlenecks and improve efficiency |
| 🛡️ Harden | 06_harden/ | 4 | Secure, monitor, and make resilient |
| 🔬 Review | 07_review/ | 6 | Architect, test, and future-proof |
📁 Repository Structure
senior-engineer-skills/
├── README.md
├── LICENSE
├── templates/
│ └── skill-template.md
└── skills/
├── 01_understand/
│ ├── map-code-flow.md
│ ├── identify-risky-areas.md
│ ├── dependency-analysis.md
│ └── explain-business-logic.md
├── 02_debug/
│ ├── reproduce-bug-from-logs.md
│ ├── trace-execution-path.md
│ ├── detect-race-conditions.md
│ └── analyze-performance-bottlenecks.md
├── 03_fix/
│ ├── patch-critical-bug.md
│ ├── edge-case-handling.md
│ ├── input-validation-fix.md
│ └── error-handling-fix.md
├── 04_refactor/
│ ├── extract-function-module.md
│ ├── reduce-complexity.md
│ ├── rename-for-clarity.md
│ └── introduce-design-patterns.md
├── 05_optimize/
│ ├── optimize-queries.md
│ ├── frontend-performance.md
│ ├── memory-cpu-efficiency.md
│ └── caching-strategies.md
├── 06_harden/
│ ├── security-review.md
│ ├── add-logging-monitoring.md
│ ├── validation-schema-enforcement.md
│ └── fail-safe-retry-logic.md
└── 07_review/
├── architecture-review.md
├── automated-test-generation.md
├── refactor-legacy-code-incrementally.md
├── tradeoff-analysis.md
├── code-style-consistency.md
└── future-proofing-analysis.md
🚀 How to Use
With Claude Code
Copy any skill file's content into your Claude Code session as a CLAUDE.md skill, or reference it directly:
cat skills/02_debug/trace-execution-path.md >> CLAUDE.md
As a System Prompt
Use the skill's Instructions section as a system prompt when starting a new Claude conversation focused on that task.
In a Pipeline
Chain skills together for end-to-end workflows:
1. map-code-flow → understand the system
2. identify-risky-areas → find where to focus
3. reproduce-bug → confirm the problem
4. trace-execution-path → understand the root cause
5. patch-critical-bug → apply the fix
6. security-review → verify no new vulnerabilities
7. automated-test-generation → lock in the fix
💡 Example Usage
Scenario: You inherited a broken payment service
Step 1 → skills/01_understand/explain-business-logic.md
"What does this payment flow actually do?"
Step 2 → skills/02_debug/reproduce-bug-from-logs.md
"Here are the Stripe webhook logs from the failed transactions..."
Step 3 → skills/02_debug/detect-race-conditions.md
"There's a double-charge happening under load..."
Step 4 → skills/03_fix/patch-critical-bug.md
"Apply the idempotency fix without touching unrelated code."
Step 5 → skills/06_harden/fail-safe-retry-logic.md
"Add retry logic with backoff for transient Stripe failures."
Step 6 → skills/07_review/automated-test-generation.md
"Generate tests covering the race condition and edge cases."
✨ Why This Is Different
Each skill encodes:
- Constraints — what NOT to do (no unnecessary rewrites, preserve behavior, minimal blast radius)
- Senior-level framing — how to think about the problem, not just how to execute
- Structured outputs — so you always know what you're getting back
- Real-world focus — edge cases, concurrency, production failures, legacy code
- Explicit reasoning — show your work, not just conclusions
- Built-in verification — self-check before delivering results
- Calibrated confidence — know when to trust the output and when to verify
📋 Skill Anatomy
Every skill follows a consistent structure:
# Skill: [Name]
## Purpose
What problem this solves and when to use it
## Reasoning Protocol ← NEW: Structured thinking before output
<reasoning>
1. Observation → 2. Pattern Recognition → 3. Hypothesis → 4. Evidence → 5. Conclusion
</reasoning>
## Instructions
Step-by-step execution guide with constraints
## Inputs
What the skill needs to work
## Output
Exact shape of what you'll get back
## Self-Verification Checklist ← NEW: Quality gate before delivery
- [ ] Evidence grounded
- [ ] No speculation as fact
- [ ] Edge cases considered
- [ ] Assumptions explicit
## Confidence Assessment ← NEW: Know when to trust output
🟢 High (90%+) | 🟡 Medium (60-90%) | 🔴 Low (<60%)
## Adversarial Self-Review ← NEW: Challenge your own analysis
What did I miss? What would a skeptic challenge?
🤝 Contributing
- Fork the repository
- Copy
templates/skill-template.md - Fill in all sections completely — no vague instructions
- Include all advanced sections: Reasoning Protocol, Self-Verification, Confidence Assessment, Adversarial Review
- Place it in the appropriate stage folder
- Open a PR with: skill name, purpose, and the problem it solves
Quality bar: Your skill should read like it was written by a principal engineer who has felt the pain this skill addresses. Every claim must be traceable to evidence.
📄 License
MIT — use freely, attribute if you share.