MCP HubMCP Hub
Volver a habilidades

security-audit-codebase

pjt222
Actualizado 2 days ago
7 vistas
17
2
17
Ver en GitHub
Desarrolloapi

Acerca de

Esta habilidad realiza auditorías de seguridad automatizadas en bases de código, detectando secretos expuestos, dependencias vulnerables, fallos de inyección y problemas del Top 10 de OWASP. Está diseñada para usarse antes del despliegue, durante revisiones periódicas o al prepararse para auditorías de cumplimiento. La herramienta opera con capacidades de lectura, escritura, edición y bash para escanear sistemáticamente los archivos del proyecto.

Instalación rápida

Claude Code

Recomendado
Principal
npx skills add pjt222/agent-almanac -a claude-code
Comando PluginAlternativo
/plugin add https://github.com/pjt222/agent-almanac
Git CloneAlternativo
git clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/security-audit-codebase

Copia y pega este comando en Claude Code para instalar esta habilidad

Documentación

Security Audit Codebase

Perform a systematic security review of a codebase to identify vulnerabilities and exposed secrets.

When to Use

  • Before publishing or deploying a project
  • Periodic security review of existing projects
  • After adding authentication, API integration, or user input handling
  • Before open-sourcing a private repository
  • Preparing for a security compliance audit

Inputs

  • Required: Codebase to audit
  • Optional: Specific focus area (secrets, dependencies, injection, auth)
  • Optional: Compliance framework (OWASP, ISO 27001, SOC 2)
  • Optional: Previous audit findings for comparison

Procedure

Step 1: Scan for Exposed Secrets

Search for patterns that indicate hardcoded secrets:

# API keys and tokens
grep -rn "sk-\|ghp_\|gho_\|github_pat_\|hf_\|AKIA" --include="*.{md,js,ts,py,R,json,yml,yaml}" .

# Generic secret patterns
grep -rn "password\s*=\s*['\"]" --include="*.{js,ts,py,R,json}" .
grep -rn "api[_-]key\s*[=:]\s*['\"]" --include="*.{js,ts,py,R,json}" .
grep -rn "secret\s*[=:]\s*['\"]" --include="*.{js,ts,py,R,json}" .

# Connection strings
grep -rn "postgresql://\|mysql://\|mongodb://" .

# Private keys
grep -rn "BEGIN.*PRIVATE KEY" .

Got: No real secrets found — only placeholders like YOUR_TOKEN_HERE or [email protected].

If fail: If real secrets are found, remove them immediately, rotate the exposed credential, and clean git history with git filter-branch or git-filter-repo. Treat any exposed secret as compromised.

Step 2: Check .gitignore Coverage

Verify sensitive files are excluded:

# Check that these are git-ignored
git check-ignore .env .Renviron credentials.json node_modules/

# Look for tracked sensitive files
git ls-files | grep -i "\.env\|\.renviron\|credentials\|secret"

Got: All sensitive files (.env, .Renviron, credentials.json) are listed in .gitignore, and git ls-files returns no tracked sensitive files.

If fail: If sensitive files are tracked, run git rm --cached <file> to untrack them, add to .gitignore, and commit. The file remains on disk but is no longer version-controlled.

Step 3: Audit Dependencies

Node.js:

npm audit
npx audit-ci --moderate

Python:

pip-audit
safety check

R:

# Check for known vulnerabilities in packages
# No built-in tool, but verify package sources
renv::status()

Got: No high or critical vulnerabilities in dependencies. Moderate and low vulnerabilities documented for review.

If fail: If critical vulnerabilities are found, update the affected packages immediately with npm audit fix or pip install --upgrade. If updates introduce breaking changes, document the vulnerability and create a remediation plan.

Step 4: Check for Injection Vulnerabilities

SQL Injection:

# Look for string concatenation in queries
grep -rn "paste.*SELECT\|paste.*INSERT\|paste.*UPDATE\|paste.*DELETE" --include="*.R" .
grep -rn "query.*\+.*\|query.*\$\{" --include="*.{js,ts}" .

All database queries should use parameterized queries, not string concatenation.

Command Injection:

# Look for shell execution with user input
grep -rn "system\(.*paste\|exec(\|spawn(" --include="*.{R,js,ts,py}" .

XSS (Cross-Site Scripting):

# Look for unescaped user content in HTML
grep -rn "innerHTML\|dangerouslySetInnerHTML\|v-html" --include="*.{js,ts,jsx,tsx,vue}" .

Got: No SQL, command, or XSS injection vectors found. All database queries use parameterized statements, shell commands avoid user-controlled input, and HTML output is properly escaped.

If fail: If injection vulnerabilities are found, replace string concatenation in queries with parameterized queries, sanitize or escape user input before shell execution, and use framework-safe rendering methods instead of innerHTML or dangerouslySetInnerHTML.

Step 5: Review Authentication and Authorization

Checklist:

  • Passwords hashed with bcrypt/argon2 (not MD5/SHA1)
  • Session tokens are random and sufficiently long
  • Authentication tokens have expiration
  • API endpoints check authorization
  • CORS configured restrictively
  • CSRF protection enabled for state-changing operations

Got: All checklist items pass: passwords use strong hashing, tokens are random with expiration, endpoints enforce authorization, CORS is restrictive, and CSRF protection is active.

If fail: Prioritize fixes by severity: weak password hashing and missing authorization are critical, while CORS and CSRF issues are high. Document all findings with their severity level.

Step 6: Check Configuration Security

# Debug mode in production configs
grep -rn "debug\s*[=:]\s*[Tt]rue\|DEBUG\s*=\s*1" --include="*.{json,yml,yaml,toml,cfg}" .

# Permissive CORS
grep -rn "Access-Control-Allow-Origin.*\*\|cors.*origin.*\*" --include="*.{js,ts}" .

# HTTP instead of HTTPS
grep -rn "http://" --include="*.{js,ts,py,R}" . | grep -v "localhost\|127.0.0.1\|http://"

Got: Debug mode is disabled in production configurations, CORS does not use wildcard origins in production, and all external URLs use HTTPS.

If fail: If debug mode is enabled in production configs, disable it immediately. Replace wildcard CORS origins with explicit allowed domains. Update http:// URLs to https:// where the endpoint supports it.

Step 7: Document Findings

Create an audit report:

# Security Audit Report

**Date**: YYYY-MM-DD
**Auditor**: [Name]
**Scope**: [Repository/Project]
**Status**: [PASS/FAIL/CONDITIONAL]

## Findings Summary

| Category | Status | Details |
|----------|--------|---------|
| Exposed secrets | PASS | No secrets found |
| .gitignore | PASS | Sensitive files excluded |
| Dependencies | WARN | 2 moderate vulnerabilities |
| Injection | PASS | Parameterized queries used |
| Auth/AuthZ | N/A | No authentication in scope |
| Configuration | PASS | Debug mode disabled |

## Detailed Findings

### Finding 1: [Title]
- **Severity**: Low / Medium / High / Critical
- **Location**: `path/to/file:line`
- **Description**: What was found
- **Recommendation**: How to fix
- **Status**: Open / Resolved

## Recommendations
1. Update dependencies to fix moderate vulnerabilities
2. [Additional recommendations]

Got: A complete SECURITY_AUDIT_REPORT.md saved in the project root with findings categorized by severity, each with a specific location, description, and recommendation.

If fail: With too many findings to document individually, group by category and prioritize critical/high findings. Generate the report regardless of outcome to establish a baseline.

Validation

  • No hardcoded secrets in source code
  • .gitignore covers all sensitive files
  • No high/critical dependency vulnerabilities
  • No injection vulnerabilities
  • Authentication is properly implemented (if applicable)
  • Audit report is complete and findings addressed

Pitfalls

  • Only checking current files: Secrets in git history are still exposed. Check with git log -p --all -S 'secret_pattern'.
  • Ignoring dev dependencies: Development dependencies can still introduce supply chain risks.
  • False sense of security from .gitignore: .gitignore only prevents future tracking. Already-committed files need git rm --cached.
  • Overlooking configuration files: docker-compose.yml, CI configs, and deployment scripts often contain secrets.
  • Not rotating compromised credentials: Finding and removing a secret is not enough. The credential must be revoked and regenerated.

Related Skills

  • configure-git-repository - proper .gitignore setup
  • write-claude-md - documenting security requirements
  • setup-gxp-r-project - security in regulated environments

Repositorio GitHub

pjt222/agent-almanac
Ruta: i18n/caveman-lite/skills/security-audit-codebase
0
agentsagentskillsai-assisted-developmentclaude-codeskillsteams

Habilidades relacionadas

qmd

Desarrollo

qmd es una herramienta CLI de búsqueda e indexación local que permite a los desarrolladores indexar y buscar en archivos locales mediante búsqueda híbrida que combina BM25, embeddings vectoriales y reranking. Es compatible tanto con uso desde la línea de comandos como con modo MCP (Model Context Protocol) para integración con Claude. La herramienta utiliza Ollama para los embeddings y almacena los índices localmente, lo que la hace ideal para buscar documentación o bases de código directamente desde la terminal.

Ver habilidad

subagent-driven-development

Desarrollo

Esta habilidad ejecuta planes de implementación asignando un nuevo subagente para cada tarea independiente, con revisión de código entre tareas. Permite una iteración rápida mientras mantiene controles de calidad a través de este proceso de revisión. Úsala cuando trabajes en tareas mayormente independientes dentro de la misma sesión para garantizar un progreso continuo con verificaciones de calidad integradas.

Ver habilidad

mcporter

Desarrollo

La habilidad mcporter permite a los desarrolladores gestionar y llamar servidores del Protocolo de Contexto de Modelo (MCP) directamente desde Claude. Proporciona comandos para listar servidores disponibles, llamar a sus herramientas con argumentos, y manejar la autenticación y el ciclo de vida del daemon. Utiliza esta habilidad para integrar y probar la funcionalidad de servidores MCP en tu flujo de trabajo de desarrollo.

Ver habilidad

adk-deployment-specialist

Desarrollo

Esta habilidad despliega y orquesta agentes Vertex AI ADK utilizando el protocolo A2A, gestionando el descubrimiento de AgentCard, el envío de tareas y soportando herramientas como el Sandbox de Ejecución de Código y el Banco de Memoria. Permite construir sistemas multiagente con patrones de orquestación secuencial, paralela o en bucle en Python, Java o Go. Úsela cuando se le solicite desplegar agentes ADK u orquestar flujos de trabajo de agentes en Google Cloud.

Ver habilidad