Security Audit

安全审计

SKILL.md

--- frontmatter

name: "Security Audit"
id: "security-audit"
version: "1.0"
claude_subagent: "security-auditor"
fallback: "self-check"
min_tad_version: "2.1"
platforms: ["claude", "codex", "gemini"]
conditional: true
trigger_pattern: "auth|token|password|credential|api.*key|encrypt|decrypt|session|cookie|sql|query|upload|file|exec|eval"

Security Audit Skill

Purpose

Analyze code for security vulnerabilities, potential attack vectors, and compliance with security best practices.

When to Use

•When code contains security-sensitive patterns
•During Gate 3 (implementation quality)
•During Gate 4 (integration verification)
•For authentication/authorization changes
•For data handling logic

Trigger Patterns

This skill is conditionally triggered when code matches:

regex

auth|token|password|credential|api.*key|encrypt|decrypt|session|cookie|sql|query|upload|file|exec|eval

Checklist

P0 - Blocking (Must Pass)

• No SQL injection vulnerabilities
• No command injection vulnerabilities
• No XSS (Cross-Site Scripting) vulnerabilities
• No hardcoded secrets or credentials
• No authentication bypass risks
• No sensitive data exposure in logs/errors

P1 - Critical (Must Pass)

• Proper input validation on all user inputs
• Secure session management
• Correct authorization checks
• Safe file upload handling
• Secure API key/token handling

P2 - Warning (Should Address)

• Security headers configured
• HTTPS enforced where applicable
• Rate limiting implemented
• Audit logging for sensitive operations
• Dependencies have no known CVEs

P3 - Informational (Nice-to-have)

• Security documentation updated
• Threat model considerations
• Penetration testing recommendations
• Future security improvements

Pass Criteria

Severity	Requirement
P0	Zero issues allowed - blocks release
P1	Zero issues allowed - must fix before merge
P2	Max 5 with remediation plan
P3	Optional improvements, document only

Evidence Output

Path: .tad/evidence/reviews/{date}-security-audit-{task}.md

Execution Contract

•Input: file_paths[], context{}, patterns_found[]
•Output: {passed: bool, vulnerabilities: [{severity, category, file, line, description, remediation}], evidence_path: string}
•Timeout: 240s
•Parallelizable: true

Claude Enhancement

When running on Claude Code, call subagent security-auditor for deeper analysis. Reference: .tad/templates/output-formats/security-review-format.md

Vulnerability Categories

•Injection (SQL, Command, XSS)
•Authentication/Authorization
•Data Exposure
•Cryptographic Issues
•Security Misconfiguration
•Input Validation