A deep-dive into the skill-security-auditor Claude Code blueprint — an automated security review workflow that integrates directly into your development pipeline and catches supply-chain attacks before they ship.
**TL;DR:** `skill-security-auditor` is an ARCHITECT-tier Claude Code blueprint that runs automated security reviews on code changes — catching secrets, SAST vulnerabilities, and supply-chain anomalies before they reach production. Ships in under 5 minutes.
The 10-Second Pitch
- Runs **SAST scans** (Semgrep, Bandit), **secret detection** (Gitleaks), and **dependency auditing** (npm audit, pip-audit) on every PR
- Generates a structured **JSON security report** with CVE matches, severity scores, and remediation prompts
- Tier-rated: ARCHITECT — meaning it orchestrates multiple security tools, not just one
- Designed for **shift-left security** teams that want automated review without manual intervention
Setup Directions
Step 1 — Install the Blueprint
\`\`\`bash
mrt install "skill-security-auditor"
\`\`\`
Step 2 — Connect Your CI Environment
In your project root, create `.security-auditor.json`:
\`\`\`json
{
"tools": ["gitleaks", "semgrep", "trivy"],
"fail_on": ["critical", "high"],
"report_format": "json",
"notify_slack": true
}
\`\`\`
Step 3 — Run Your First Audit
\`\`\`bash
claude -- blueprint security-auditor --target ./src --mode ci
\`\`\`
Step 4 — Review the Report
Output lands in `.security-report.json`:
\`\`\`json
{
"scanned_files": 142,
"critical_findings": 0,
"high_findings": 2,
"secrets_detected": ["AWS_SECRET_KEY"],
"cve_matches": ["CVE-2024-1234"],
"remediation": "Upgrade lodash@>=4.17.21"
}
\`\`\`
The Exact Prompt for an Ad-Hoc Review
Run a full security audit on the ./src directory. Use gitleaks for
secrets detection, semgrep for SAST, and trivy for dependency scanning.
Output results as structured JSON. Fail if any CRITICAL findings are found.
Pros & Cons
| Pros | Cons |
|---|
| Multi-tool orchestration (not just one scanner) | Requires `claude` CLI in PATH |
| Structured JSON output for automation | Some tools (Trivy) need Docker |
|---|
| ARCHITECT tier — serious workflow design | Semgrep rules must be kept up to date |
|---|
| Zero false positives on secret detection (Gitleaks) | May be noisy on large legacy codebases |
|---|
| Integrates with GitHub/GitLab PR comment workflows | Not a real-time file monitor — needs triggering |
|---|
Verdict
`skill-security-auditor` fills the gap between "security scan available" and "security scan actually runs on every PR." The ARCHITECT designation means it doesn't just call one tool — it orchestrates a pipeline of them and produces a unified report. If you're shipping code without this in your CI, you're accepting risk you don't need to.
**Best for:** DevSecOps teams, open-source maintainers, and any organization handling sensitive data in their codebase.
**Alternative:** For lightweight single-tool scanning, try `gitleaks` directly. For full SIEM integration, look at the Enterprise tier `security-auditing` blueprint instead.
*Blueprint available at [mr.technology/blueprints](/blueprints). ARCHITECT-tier skills require Pro access.*