Deep security audit of code, dependencies, and infrastructure. Checks OWASP Top 10, authentication, authorization, data protection, and dependency vulnerabilities. Use before releases or after significant changes.
Scanned 5/28/2026
Install via CLI
openskills install ArtemioPadilla/agent-triforce---
name: security-audit
description: >
Deep security audit of code, dependencies, and infrastructure. Checks OWASP Top 10,
authentication, authorization, data protection, and dependency vulnerabilities.
Use before releases or after significant changes.
context: fork
agent: centinela-qa
---
Perform a security audit on: $ARGUMENTS
If no specific scope is provided, audit the entire `src/` directory.
Follow these steps:
**SIGN IN:**
- Run the SIGN IN checklist from your agent file
- Note areas of concern based on the scope and past findings
**AUDIT:**
1. **OWASP Top 10** systematic check:
- A01: Broken Access Control — check auth/authz on all endpoints
- A02: Cryptographic Failures — check encryption, key management, hashing
- A03: Injection — SQL, NoSQL, OS command, LDAP injection vectors
- A04: Insecure Design — business logic flaws, missing rate limits
- A05: Security Misconfiguration — default configs, verbose errors, CORS
- A06: Vulnerable Components — `npm audit` / `pip audit`
- A07: Auth Failures — weak passwords, missing MFA, session management
- A08: Data Integrity — deserialization, unsigned updates
- A09: Logging Failures — sensitive data in logs, missing audit trail
- A10: SSRF — server-side request forgery vectors
2. **Secrets scan**: grep for API keys, tokens, passwords, connection strings
3. **Dependency audit**: check all deps for known CVEs
4. **Smart contracts** (if Solidity): reentrancy, integer overflow, access control, front-running
5. **Architecture review**: Check dependency direction (business logic must not depend on frameworks), verify layer separation (no auth bypass through adapter shortcuts, no direct DB access from use cases)
6. **Test strategy review**: Verify security-critical paths have tests, authentication/authorization flows are integration-tested, input validation has unit tests
### License Compliance Scan
Scan all project dependencies for license compatibility with the project license (MIT):
1. **Detect dependency manifest**: `requirements.txt`, `pyproject.toml` (Python) or `package.json` (Node)
2. **Check each dependency's license** using `pip-licenses --format=markdown` (Python) or `npx license-checker --json` (Node). If tools unavailable, inspect manifests manually
3. **Classify by compatibility**:
- **Pass** (permissive): MIT, BSD-2-Clause, BSD-3-Clause, Apache-2.0, ISC, Unlicense, CC0-1.0, 0BSD
- **Warning** (weak copyleft): LGPL-2.1, LGPL-3.0, MPL-2.0, EPL-1.0, EPL-2.0 -- flag with explanation of linking implications
- **Critical** (strong copyleft): GPL-2.0, GPL-3.0, AGPL-3.0 -- incompatible with MIT distribution
- **Critical** (no license): Dependency has no detectable license -- cannot legally redistribute
4. **Output**: License compatibility table in the security audit report with columns: Dependency, Version, License, Compatibility (Pass/Warning/Critical)
References: Skaife (2021), EPAM SolutionsHub (2023), Leroux (2025)
If any Critical finding: invoke the NON-NORMAL: Critical Vulnerability Response checklist from your agent file.
**⏸️ TIME OUT — Run Verification Checklists (DO-CONFIRM):**
7. Run through the Security Verification checklist from your agent file
8. Run through the Quality Verification checklist from your agent file
9. Issue verdict based on findings
**SIGN OUT:**
10. Write report to `docs/reviews/security-audit-{date}.md`
11. Write the Findings Handoff-to-Forja using the communication checklist
12. Run the SIGN OUT checklist from your agent file
No comments yet. Be the first to comment!