Iron Law: Never approve a security review with unresolved CRITICAL or HIGH findings; always escalate to the human before proceeding past a security gate.
Security Review Skill
Purpose
Provides security review methodology, vulnerability checklists, and remediation patterns for application code.
Process
- Load checklist -- Read the security review checklist for review categories and severity levels
- Scan code -- Use Grep/Glob to find security-sensitive patterns
- Evaluate -- Check each finding against the checklist
- Report -- Output findings with severity and remediation guidance
For the complete security review checklist and methodology:
Read reference/security-review-checklist.md
Reference Files
| File |
Content |
Load When |
| reference/security-review-checklist.md |
OWASP Top 10, secrets scanning, auth/authz, injection, data protection checklists |
All security reviews |
| reference/pci-dss-requirements.md |
PCI DSS 12 requirements, compliance levels (L1-L4), SAQ types, prohibited data, audit log requirements, common violations |
Any feature touching payment card data, payment processors, or billing |
| reference/owasp-infrastructure-baseline.md |
15 OWASP-mapped infrastructure controls — encryption at rest/transit, IAM least-privilege, network hardening, audit logging, secret rotation |
Any IaC review (Terraform, GCP, AWS), cloud config review, infrastructure security |
| reference/agent-guardrails-checklist.md |
12-layer AI agent guardrail pipeline, prompt injection defense, output validation, async audit logging, Constitutional AI — agent-specific security controls |
Any LangGraph agent, agentic AI service, or AI feature with tool use |
| reference/claude-config-security.md |
Manual security audit checklist for .claude/ config (settings.json, hooks, agents, MCP servers, CLAUDE.md) — workspace-specific allow/deny/ask list review, hook injection surface, MCP supply chain risks |
Before committing any .claude/ directory changes |
Error Handling
If target files/directories don't exist, report "Target not found" with the paths searched.
If a scan produces no findings, report "No security issues detected" with scope of scan.
1---2name: security-reviewer3description: Security vulnerability detection and remediation skill. Provides OWASP Top 10 checklists, secret scanning patterns, and security review methodology.4---56**Iron Law:** Never approve a security review with unresolved CRITICAL or HIGH findings; always escalate to the human before proceeding past a security gate.78# Security Review Skill910## Purpose1112Provides security review methodology, vulnerability checklists, and remediation patterns for application code.1314## Process15161. **Load checklist** -- Read the security review checklist for review categories and severity levels172. **Scan code** -- Use Grep/Glob to find security-sensitive patterns183. **Evaluate** -- Check each finding against the checklist194. **Report** -- Output findings with severity and remediation guidance2021For the complete security review checklist and methodology:2223Read [reference/security-review-checklist.md](reference/security-review-checklist.md)2425## Reference Files2627| File | Content | Load When |28|------|---------|-----------|29| [reference/security-review-checklist.md](reference/security-review-checklist.md) | OWASP Top 10, secrets scanning, auth/authz, injection, data protection checklists | All security reviews |30| [reference/pci-dss-requirements.md](reference/pci-dss-requirements.md) | PCI DSS 12 requirements, compliance levels (L1-L4), SAQ types, prohibited data, audit log requirements, common violations | Any feature touching payment card data, payment processors, or billing |31| [reference/owasp-infrastructure-baseline.md](reference/owasp-infrastructure-baseline.md) | 15 OWASP-mapped infrastructure controls — encryption at rest/transit, IAM least-privilege, network hardening, audit logging, secret rotation | Any IaC review (Terraform, GCP, AWS), cloud config review, infrastructure security |32| [reference/agent-guardrails-checklist.md](reference/agent-guardrails-checklist.md) | 12-layer AI agent guardrail pipeline, prompt injection defense, output validation, async audit logging, Constitutional AI — agent-specific security controls | Any LangGraph agent, agentic AI service, or AI feature with tool use |33| [reference/claude-config-security.md](reference/claude-config-security.md) | Manual security audit checklist for `.claude/` config (settings.json, hooks, agents, MCP servers, CLAUDE.md) — workspace-specific allow/deny/ask list review, hook injection surface, MCP supply chain risks | Before committing any `.claude/` directory changes |3435## Error Handling3637If target files/directories don't exist, report "Target not found" with the paths searched.38If a scan produces no findings, report "No security issues detected" with scope of scan.