shield_lock
金大哥
← 返回 Skill 列表
extension
分类: 开发与工程无需 API Key

codex-security-review

以安全为重点的代码审查，使用OWASP Top 10和CWE模式。通过静态分析检测漏洞、秘密信息、认证问题以及安全配置错误。

person作者: jakexiaohubgithub
open_in_new仓库download下载资源包

Codex Security Review

Purpose

Use this skill to perform security-focused review of code changes, identifying vulnerabilities aligned with OWASP Top 10 2021 and common CWE patterns.

Prerequisites

  • Working directory with source code
  • Optional: dependency manifest files (package.json, requirements.txt, go.mod) for supply chain analysis
  • codex CLI is installed and authenticated
  • codex-review skill pack is installed (npx github:lploc94/codex_skill)

Runner

RUNNER="{{RUNNER_PATH}}"

Workflow

  1. Ask user to choose review effort level: low, medium, high, or xhigh (default: high). Ask review scope: working-tree (uncommitted changes), branch (branch diff), or full (entire codebase). Set EFFORT and SCOPE.
  2. Build prompt from references/prompts.md (Security Review Prompt with OWASP checklist).
  3. Start round 1 with node "$RUNNER" start --working-dir "$PWD" --effort "$EFFORT".
  4. Poll with adaptive intervals (Round 1: 60s/60s/30s/15s..., Round 2+: 30s/15s...). After each poll, report specific activities from poll output (e.g. which files Codex is analyzing, what vulnerability patterns it's checking). See references/workflow.md for parsing guide. NEVER report generic "Codex is running" — always extract concrete details.
  5. Parse security findings with references/output-format.md (includes CWE/OWASP mappings).
  6. Fix valid vulnerabilities in code; rebut false positives with evidence.
  7. Resume debate via --thread-id until APPROVE or stalemate.
  8. Return final security assessment with risk summary.

Effort Level Guide

| Level | Depth | Best for | |----------|-------------------|---------------------------------| | low | Common patterns | Quick security sanity check | | medium | OWASP Top 10 | Standard security review | | high | Deep analysis | Pre-production security audit | | xhigh | Exhaustive | Critical/regulated systems |

Scope Guide

| Scope | Coverage | Best for | |----------------|------------------------------------|-----------------------------| | working-tree | Uncommitted changes only | Pre-commit security check | | branch | Branch diff vs base | Pre-merge security review | | full | Entire codebase | Security audit |

Security Categories Covered

OWASP Top 10 2021

  • A01:2021 - Broken Access Control
  • A02:2021 - Cryptographic Failures
  • A03:2021 - Injection (SQL, Command, XSS, etc.)
  • A04:2021 - Insecure Design
  • A05:2021 - Security Misconfiguration
  • A06:2021 - Vulnerable and Outdated Components
  • A07:2021 - Identification and Authentication Failures
  • A08:2021 - Software and Data Integrity Failures
  • A09:2021 - Security Logging and Monitoring Failures
  • A10:2021 - Server-Side Request Forgery (SSRF)

Additional Security Checks

  • Secrets/credentials in code
  • Hardcoded passwords and API keys
  • Insecure random number generation
  • Path traversal vulnerabilities
  • XML External Entity (XXE) attacks
  • Insecure deserialization
  • Missing security headers
  • CORS misconfigurations

Output Format

Each security finding includes:

  • CWE ID: Common Weakness Enumeration identifier
  • OWASP Category: OWASP Top 10 2021 mapping
  • Severity: critical, high, medium, low
  • Confidence: high, medium, low (static analysis confidence)
  • Attack Vector: How the vulnerability could be exploited
  • Suggested Fix: Secure code example

See references/output-format.md for complete specification.

Important Limitations

This is static analysis only:

  • ✅ Can detect: Code patterns, hardcoded secrets, common vulnerabilities
  • ❌ Cannot detect: Runtime behavior, memory leaks (need profiling), zero-days
  • ⚠️ Heuristic: Findings are AI-generated suggestions, not guaranteed vulnerabilities

Always:

  • Verify findings manually before treating as confirmed vulnerabilities
  • Run dynamic security testing (DAST) for runtime issues
  • Use dedicated tools for dependency scanning (Snyk, Dependabot)
  • Consult security experts for critical systems

Required References

  • Detailed execution: references/workflow.md
  • Prompt templates: references/prompts.md
  • Output contract: references/output-format.md

Rules

  • Codex reviews only; it does not edit files
  • Mark all findings with confidence level (high/medium/low)
  • Provide CWE and OWASP mappings for all vulnerabilities
  • Include attack vector explanation for each finding
  • If stalemate persists, present both sides and defer to user
  • Never claim 100% security coverage - static analysis has limits