Verified Agent Skill record
Skill Security Auditor
A guide for security auditing of source code, dependencies, and configurations covering secrets detection, OWASP Top 10 review, language-specific vulnerability checks for Python, JavaScript/TypeScript, and Go, dependency audits, configuration review, and authentication/authorization review, with a markdown report format.
Tasks
- Perform reconnaissance on the target project to identify languages, frameworks, entry points (routes, controllers, API handlers), authentication handling, external service calls, and where user input is accepted
- Scan a project directory for hardcoded secrets such as API keys, tokens, database connection strings with embedded passwords, committed private keys or certificates, plaintext secrets in .env or config files, and secrets in CI/CD configuration
- Review source code against the OWASP Top 10 checklist (broken access control, cryptographic failures, injection, insecure design, security misconfiguration, vulnerable components, authentication failures, data integrity failures, logging failures, SSRF)
- Apply language-specific manual code review checks for Python (SQL injection, command injection, path traversal), JavaScript/TypeScript (XSS via innerHTML, prototype pollution, eval), and Go (SQL injection, path traversal), following the dangerous/safe examples in the guide
- Audit project dependencies for known CVEs using available tools (pip audit, safety, npm audit, govulncheck, or trivy if installed), noting that tool availability is not guaranteed
- Review configuration files for insecure defaults such as debug mode in production, wildcard ALLOWED_HOSTS, open CORS, default secret keys, and disabled TLS verification, plus Dockerfiles, CI/CD, and cloud configs
- Review authentication and authorization: password hashing algorithms, session timeouts and rotation, JWT validation, endpoint authorization checks, RBAC consistency, and rate limiting on login and sensitive endpoints
- Generate a security audit report using the documented markdown structure with summary, findings (severity, category, location, description, impact, recommendation), and statistics
Inputs
- A project directory containing source code to audit
- Dependency manifests such as package.json, pyproject.toml, requirements.txt, go.mod, or pom.xml
- Configuration files including application configs, Dockerfiles, CI/CD pipelines, and cloud configurations
Outputs
- A security audit report in markdown containing a summary (project, date, scope, risk level), findings with severity, OWASP category, location, description, impact, and recommendation, and statistics by severity
- Lists of detected secrets and insecure configuration findings
Limitations and checks
- The referenced scripts/ and references/ directories are not verified to exist; only SKILL.md is evidenced, so running the bundled scan scripts may not be possible
- The scan described for scan_project.py is lightweight, covering hardcoded secrets, dangerous function calls, and insecure patterns; it does not constitute full vulnerability scanning, which requires the manual workflows
- No guarantee that any dependency audit tool is installed; severity output formats vary across the listed tools, so severity categorization may require manual normalization
- No specific platform support, installation method, or automated remediation is documented
- Confirm the project's languages, frameworks, and entry points were identified before reviewing code, per the reconnaissance step
- Verify each finding in the report includes severity, OWASP category, location (file:line), description, impact, and recommendation
- Confirm secrets scanning covered source files, config and .env files, and CI/CD configuration; if the scan script is unavailable, perform the pattern checks manually per the guide's list
- Confirm that dependency audit results reflect only tools actually installed, and normalize or manually assign severity where a tool's output does not provide it
Search
Find projects, verify facts, compare options, and filter by your requirements
Try a searchA click only fills the search box; you stay in control
Project Details
0