Security Reviewer gives the agent an audit workflow: confirm written scope and authorization, run SAST, dependency and secrets tools, review authentication, input handling and crypto by hand, classify findings by CVSS severity and report them with location, impact and remediation.
The skill names the commands to run (semgrep, bandit, gitleaks, npm audit, trivy) and includes a finding template with ID, severity, file and line, description, impact, remediation and CWE references.
What You Get
- A five-step audit workflow with an authorization check before active testing
- Example scan commands for Semgrep, Bandit, Gitleaks, npm audit and Trivy
- A finding template with ID, severity, file and line, impact, remediation and references
- An output layout: executive summary, findings table, detailed findings and prioritized recommendations
- Reference files for SAST tools, vulnerability patterns, secret scanning, penetration testing, infrastructure security and the report template
Common Use Cases
- Reviewing a pull request or a whole repository for injection and auth flaws
- Scanning a repository for committed secrets
- Auditing dependencies and container images for known vulnerabilities
- Producing a prioritized report for a compliance review
What Makes It Unique
It treats scope and authorization as part of the workflow: written authorization is confirmed before any active testing, exploits stop at proof of concept, and critical findings are reported at once. It is a reviewer that reports findings. The sibling Secure Code Guardian writes the fixes while building features.
Under The Hood
The skill is a SKILL.md with six reference files that the agent loads by topic. It ships no scripts. Its frontmatter allows only Read, Grep and Glob, which are read-only file tools, yet the instructions tell the agent to run shell commands such as semgrep, bandit, gitleaks, npm audit and trivy against your project, so those commands need shell access you grant separately. The references also cover active penetration testing and reconnaissance with tools such as nmap, which send network traffic to the target, and the skill requires written authorization before that step.