fleet-security-audit
How to conduct a security review of code or infrastructure
SKILL.md.claude/skills/fleet-security-audit/SKILL.md
GitHubFleet Security Audit Skill
You are conducting a security review.
1. Scope
- What are you reviewing? (PR, deployment, configuration, dependencies)
- What project? What component?
2. Check Areas
Code Review:
- No secrets in code (tokens, keys, passwords)
- No hardcoded credentials
- Input validation at boundaries
- No SQL injection / command injection patterns
- No XSS patterns in output
- Dependencies checked for known CVEs
- No world-writable permissions (chmod 777)
Infrastructure Review:
- Default credentials changed
- Ports properly restricted
- Network isolation between services
- TLS/encryption for external connections
- Secrets in .env (gitignored), not in code
- Container images from trusted registries
Supply Chain:
- Dependencies from official sources
- No unnecessary dependencies
- Lock files present and up to date
3. Report Findings
fleet_alert(
severity="critical|high|medium|low",
title="Finding title",
details="What was found + evidence + recommendation",
category="security"
)
4. Security Hold (Critical)
For critical findings, request security hold:
fleet_notify_human(
title="Security hold: {finding}",
message="Details + recommendation",
priority="urgent"
)