convez.ai
SkillSegurançaIntermediário
markusweldon

OWASP Security Skills para Claude Code

markusweldon/claude-owasp-security-skills

Padrões OWASP automáticos integrados ao seu workflow de código seguro

Ver no GitHub
InicialPluginScore · de 0 a 100

Conjunto de 5 skills especializadas que integram os padrões OWASP atualizados (Top 10:2025, API, Mobile, LLM/Agentic e CI/CD) diretamente no fluxo de trabalho do Claude Code. Ativa automaticamente durante desenvolvimento web, APIs, mobile, chatbots e pipelines CI/CD, fornecendo contexto de segurança em tempo real sem necessidade de memorizar diretrizes.

Instalação

$npx convez-ai markusweldon/claude-owasp-security-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
57%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Revisar código web com proteção contra OWASP Top 10:2025
  • Desenvolver APIs REST/GraphQL com prevenção de SSRF e rate limiting
  • Construir aplicativos mobile iOS/Android com conformidade OWASP
  • Implementar agentes IA e RAG com mitigação de riscos LLM
  • Securizar pipelines CI/CD com verificações IaC e dependências

Veredito

Solução prática e bem-estruturada que traz rigor OWASP para o desenvolvimento sem overhead cognitivo. A ativação automática por contexto e o design compacto (5k tokens) são diferenciais reais para equipes que precisam manter segurança sem sacrificar velocidade.

README

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

OWASP Security Skills for Claude Code

Bring the latest OWASP security standards directly into your Claude Code workflow — automatically, for every language and platform you work on.

License: MIT OWASP Top 10 OWASP ASVS OWASP LLM OWASP Agentic OWASP API OWASP Mobile OWASP CI/CD


What This Is

Five focused Claude Code skills that activate automatically when you're doing relevant work — giving Claude the full context of current OWASP standards without you having to remember to ask.

Each skill is purpose-built and stays within Claude Code's compaction window (~5,000 tokens) so guidance survives long sessions.

SkillStandardsActivates When...
owasp-webTop 10:2025, ASVS 5.0Reviewing web app code, auth, input handling, 20+ languages
owasp-apiAPI Security Top 10:2023Building REST/GraphQL APIs, SSRF prevention, JWT, rate limiting
owasp-mobileMobile Top 10:2024iOS/Android/React Native/Flutter development
owasp-llmLLM Top 10:2025, Agentic:2026Building chatbots, RAG, agents, MCP servers, function calling
owasp-cicdCI/CD Security Top 10GitHub Actions, Dockerfiles, IaC, dependency management

Quick Install

Project-level (recommended for team use)

# All 5 skills
curl -sL https://raw.githubusercontent.com/markusweldon/claude-owasp-security-skills/main/install.sh | bash

# Or install individual skills:
curl -sL https://raw.githubusercontent.com/markusweldon/claude-owasp-security-skills/main/.claude/skills/owasp-web/SKILL.md \
  -o .claude/skills/owasp-web/SKILL.md --create-dirs

curl -sL https://raw.githubusercontent.com/markusweldon/claude-owasp-security-skills/main/.claude/skills/owasp-api/SKILL.md \
  -o .claude/skills/owasp-api/SKILL.md --create-dirs

curl -sL https://raw.githubusercontent.com/markusweldon/claude-owasp-security-skills/main/.claude/skills/owasp-llm/SKILL.md \
  -o .claude/skills/owasp-llm/SKILL.md --create-dirs

curl -sL https://raw.githubusercontent.com/markusweldon/claude-owasp-security-skills/main/.claude/skills/owasp-mobile/SKILL.md \
  -o .claude/skills/owasp-mobile/SKILL.md --create-dirs

curl -sL https://raw.githubusercontent.com/markusweldon/claude-owasp-security-skills/main/.claude/skills/owasp-cicd/SKILL.md \
  -o .claude/skills/owasp-cicd/SKILL.md --create-dirs

Global install (all your projects)

curl -sL https://raw.githubusercontent.com/markusweldon/claude-owasp-security-skills/main/install.sh | bash -s -- --global

Clone and install manually

git clone https://github.com/markusweldon/claude-owasp-security-skills.git
cd claude-owasp-security-skills
bash install.sh                    # project-level
bash install.sh --global           # global (~/.claude/skills/)
bash install.sh --skill owasp-llm  # single skill only

What Each Skill Covers

owasp-web — Web Application Security

  • OWASP Top 10:2025 quick reference + prevention patterns
  • Security code review checklist (input, auth, access control, data, errors)
  • Secure vs unsafe code patterns for SQL injection, command injection, password storage, access control, fail-closed error handling
  • ASVS 5.0 verification levels (L1/L2/L3)
  • Language-specific security pitfalls for JavaScript, Python, Java, C#, PHP, Go, Ruby, Rust, C/C++, Shell, SQL (10 languages in skill; 10 more in docs/)
  • Deep security analysis mindset checklist

owasp-api — API Security

  • OWASP API Security Top 10:2023 full coverage
  • BOLA/IDOR prevention, mass assignment mitigation
  • SSRF prevention with blocked internal IP ranges
  • JWT security: algorithm pinning, claims validation, storage
  • GraphQL-specific risks: depth limits, complexity, introspection
  • Rate limiting patterns, API versioning & inventory checklist

owasp-mobile — Mobile Security

  • OWASP Mobile Top 10:2024 full coverage
  • iOS: Keychain storage, ATS configuration, certificate pinning, jailbreak detection, screen masking
  • Android: Keystore/EncryptedSharedPreferences, manifest hardening, SafetyNet/Play Integrity, ProGuard
  • Flutter/Dart: flutter_secure_storage, secure communications
  • React Native: AsyncStorage pitfalls, Hermes debugger, deep link validation
  • Certificate pinning code for URLSession and OkHttp
  • Privacy controls: PII in logs, background screen masking, permissions

owasp-llm — LLM & AI Agent Security

  • OWASP LLM Top 10:2025: prompt injection, output handling, excessive agency, system prompt leakage, vector store isolation, unbounded consumption
  • OWASP Agentic AI 2026: goal hijacking, tool misuse, inter-agent communication, cascading failures, rogue agents
  • Prompt injection defense patterns with structured roles
  • Safe LLM output handling before SQL/shell/HTML/tool calls
  • RAG/vector store tenant isolation
  • PII scrubbing before LLM calls
  • Sandboxed code execution for agent-generated code
  • Agent authentication and kill switch patterns

owasp-cicd — CI/CD & Supply Chain Security

  • OWASP CI/CD Security Top 10 full coverage
  • GitHub Actions hardening: SHA-pinned actions, minimal permissions: blocks, OIDC vs static secrets
  • Docker security: pinned digests, non-root users, multi-stage builds
  • Dependency pinning with hash verification (npm ci, --require-hashes)
  • Artifact signing with Sigstore/cosign + SBOM generation
  • Branch protection configuration reference
  • Terraform/IaC security: tfsec/checkov, state file encryption, no secrets in .tfvars
  • Webhook HMAC validation, third-party service governance

Usage Examples

Once installed, skills activate automatically. You don't need to invoke them explicitly.

"Review this Express.js authentication middleware for security issues"
→ owasp-web activates

"Is this GraphQL resolver vulnerable to BOLA?"
→ owasp-api activates

"Check my GitHub Actions workflow for security problems"
→ owasp-cicd activates

"Is my RAG pipeline safe from prompt injection?"
→ owasp-llm activates

"Review this Swift Keychain implementation"
→ owasp-mobile activates

You can also invoke skills explicitly:

/owasp-web
/owasp-api
/owasp-mobile
/owasp-llm
/owasp-cicd

Standards Coverage

StandardVersionReleased
OWASP Top 102025April 2025
OWASP ASVS5.0.0May 2025
OWASP Top 10 for LLM Applications20252025
OWASP Top 10 for Agentic Applications2026December 2025
OWASP API Security Top 1020232023
OWASP Mobile Top 1020242024
OWASP Top 10 CI/CD Security Risks20222022

Design: Why 5 Skills Instead of 1?

Claude Code compresses skills under context pressure, retaining only the first ~5,000 tokens of each active skill. A single large SKILL.md would lose its tail during long sessions — exactly when security guidance matters most.

Five focused skills means:

  • Each skill fits within the 5,000-token retention window
  • Only relevant skills activate (mobile skill won't fire during backend work)
  • Skills can be installed selectively — install just owasp-llm for an AI project
  • Descriptions stay within the 1,536-character per-skill limit for accurate triggering

Repository Structure

.claude/skills/
├── owasp-web/SKILL.md      # Web app security (Top 10:2025, ASVS 5.0)
├── owasp-api/SKILL.md      # API security (API Top 10:2023)
├── owasp-mobile/SKILL.md   # Mobile security (Mobile Top 10:2024)
├── owasp-llm/SKILL.md      # LLM + agentic AI security
└── owasp-cicd/SKILL.md     # CI/CD + supply chain security

docs/
└── language-security-reference.md  # Extended language quirks (10+ more languages)

install.sh                  # Installer script

Contributing

Contributions welcome — especially:

  • Updates when new OWASP standards are released
  • Language-specific security patterns not yet covered
  • Real-world vulnerable/safe code pattern additions
  • Bug reports for incorrect guidance

See CLAUDE.md for contributor guidelines and the development workflow.

To contribute:

  1. Fork the repository
  2. Create a feature branch: git checkout -b feat/owasp-api-graphql-patterns
  3. Make changes following the patterns in existing SKILL.md files
  4. Submit a pull request using the PR template

Sources


License

MIT — see LICENSE.


Inspired by agamm/claude-code-owasp. Enhanced with API, Mobile, CI/CD coverage, multi-skill architecture, and 2024/2025/2026 standards updates.

Ver o original no GitHub

Bom para

Relacionados