agentic-security-audit
por kingrubic
Web & Frontend DevelopmentResumen
Audit codebases, infrastructure, AND agentic AI systems for security issues.
Descripción
Qué hace esta skill
Audits codebases, infrastructure, and AI agent workspaces for security vulnerabilities. Covers dependency scanning, secret detection, OWASP web top 10 patterns, SSL/TLS verification, and file permissions. Also addresses agentic-specific risks including prompt injection, identity spoofing, memory poisoning, and multi-agent communication security based on OWASP Agentic Top 10 (2026).
Combines traditional code and infrastructure security checks with agentic AI-specific audits in one skill, covering attack surfaces that standard security tools miss entirely.
Cuándo usarla
- Scanning Node.js or Python dependencies for known CVEs before a release
- Detecting hardcoded API keys or credentials committed to a git repository
- Auditing an AI agent workspace for prompt injection vectors in agent-facing content
- Verifying SSL certificate validity and checking for disabled SSL verification in code
- Reviewing agent identity verification rules and access boundaries before exposing an agent to external interactions
Flujo de trabajo de ejemplo
User asks: 'Run a full security audit on my project'
- 1
Runs dependency vulnerability scan using npm audit, pip-audit, or govulncheck depending on project type
- 2
Searches source files for hardcoded secrets, API keys, and credentials using grep patterns against git history and current files
- 3
Checks for OWASP top 10 patterns including SQL injection, XSS, CORS misconfiguration, and disabled SSL verification
- 4
Verifies SSL certificate validity and supported TLS versions on any configured endpoints
- 5
If agent workspace files are present, runs agentic audit checking for prompt injection patterns, missing authorized-sender rules, and override instructions in memory files
Report listing critical issues and warnings by category with exit code 1 if critical issues are found
Instalación
clawhub install kingrubic/agentic-security-auditEjecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.