AgentSkillsCN

security-audit

聚焦防御性编程、漏洞识别与“黑盒”思维，展开深度安全分析。适用于敏感功能的开发，或进行全面的安全审计。

SKILL.md
--- frontmatter
name: security-audit
description: Deep-dive security analysis focusing on defensive coding, vulnerability identification, and "black-box" thinking. Use for sensitive features or comprehensive audits.

Security Audit Skill

This skill focuses on a defensive, "black-hat" perspective to secure the application. It goes beyond basic linting to identify how an attacker might exploit the system.

Mindset: "How would I break this?"

When auditing, adopt the persona of an ethical researcher looking for bounties or exploits.

Audit Checklist

1. Data Flow Analysis

  • •Entry Points: Where does data enter? (API endpoints, forms, URL params, headers).
  • •Trust Boundaries: Is trusted, internal data mixed with untrusted, external data?
  • •Sanitization: Is data validated immediately upon entry?

2. Common Vulnerabilities (OWASP Top 10 Focus)

  • •Injection: SQL, NoSQL, Command Injection?
  • •Broken Auth: Can session tokens be stolen? Are passwords salted/hashed?
  • •Sensitive Data: Are logs leaking PII or secrets? (Check console.log carefully).
  • •SDE (Insecure Direct Object References): Can User A access User B's resource by changing an ID?

3. Logic Flaws

  • •Race Conditions: Can simultaneous requests corrupt state?
  • •Business Logic Bypass: Can a user skip a payment step? Can a negative quantity be ordered?

4. Dependency Check

  • •Are we using outdated libraries with known CVEs?
  • •Are we importing massive packages for simple functions?

"Black-Box" Testing Strategy

Suggest specific tests the user can run:

  • •"Try sending a JSON payload with a circular reference."
  • •"Attempt to bypass the frontend validation using curl."
  • •"Check if error messages reveal stack traces or database structure."

Output Format

🛡️ Security Report

High Severity (Immediate Action Required)

  • •[Vulnerability Name]: Description of the exploit vector.
  • •Fix: Concrete code change to mitigate.

Medium Severity (Address Soon)

  • •[Vulnerability Name]: Risks closer to theoretical or hard to exploit.

Low Severity / Best Practice

  • •Hardening suggestions (e.g., headers, rate limiting).