This commit establishes the foundation for a comprehensive security audit tool for macOS developer machines. It audits all installed software, detects vulnerabilities via OSV.dev, and provides Claude-powered mitigation recommendations. ## Architecture - System inventory collection (Homebrew, pip, npm, Go, Rust, apps) - CVE scanning via OSV.dev API (no auth, unlimited rate limits) - Claude integration (CLI or SDK) for mitigation analysis - JSON + HTML report generation - Full test suite (34+ tests, 7 benchmarks) - Production-grade linting (11 linters via golangci-lint) - Vulnerability scanning (govulncheck) - GitHub Actions CI/CD pipeline ## Key Components - cmd/audit/: CLI entry point - internal/inventory/: System inventory parsing - internal/security/: OSV.dev querying, Claude integration, audit logic - internal/report/: JSON and HTML report generation - scripts/collect-inventory.sh: Bash script for system enumeration ## MVP Features - Homebrew package scanning - Python/Node/Go/Rust ecosystem scanning - CVSS-based severity filtering - Claude-powered recommendations (update/replace/protect/monitor) - Makefile automation (test, lint, vulnerability checks) - Pre-commit hooks configuration ## Testing & Quality - 34+ unit tests with table-driven patterns - 7 benchmark tests for performance - 11 configured linters (staticcheck, gosec, revive, etc.) - Code coverage reporting - GitHub Actions CI (tests on Go 1.22 & 1.23) - Pre-commit hook framework ## Documentation - README.md: Full feature and usage documentation - QUICKSTART.md: 3-minute setup guide - TESTING.md: Testing and code quality guide - QA_SETUP.md: Analysis tooling reference - DATA_SOURCES_SPEC.md: Vulnerability data source documentation ## Data Sources - OSV.dev: Primary CVE database (1.8 day latency, no auth needed) - GitHub Advisories: Supplement for maintainer-created advisories - OpenSSF Scorecard: Trust signals (repo health/practices) - Homebrew Formulae API: Package metadata - CISA KEV: Active exploitation tracking ## Next Steps Phase 1 (MVP): Complete and tested ✓ Phase 2 (planned): Third-party app scanning, VirusTotal integration Phase 3 (planned): Historical monitoring, scheduled audits, webhooks Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
3.1 KiB
3.1 KiB
Quick Start Guide
Get a security audit of your Mac in 3 commands.
Prerequisites
- macOS 11+
- Go 1.22+ (for development) or compiled binary
- Claude Code (for Claude integration) or Anthropic API key
Install Claude Code (optional)
For Claude analysis recommendations:
# Option 1: Using Claude Code CLI
brew install anthropic/claude/claude
# Option 2: Using Anthropic SDK
export ANTHROPIC_API_KEY=sk-ant-...
Run Audit
Option A: One Command (if you have make)
cd ~/Projects/package-review
make audit-full
open audit_report.html
Option B: Step by Step
cd ~/Projects/package-review
# 1. Collect system inventory
./scripts/collect-inventory.sh
# Output: inventory.json
# 2. Build the audit tool
go mod download
go build -o bin/audit ./cmd/audit
# 3. Run the audit
./bin/audit
# Output: audit_report.json, audit_report.html
# 4. Review the report
open audit_report.html
Interpreting Results
Summary Metrics
- Total Packages: All software scanned
- With CVEs: How many packages have known vulnerabilities
- HIGH: CVSS 7.0-8.9 (action recommended)
- CRITICAL: CVSS 9.0+ (urgent action required)
For Each Finding
Recommendation (from Claude):
- UPDATE — Upgrade to a patched version
- REPLACE — Switch to a safer alternative package
- PROTECT — Add extra protections without updating
- MONITOR — Watch for patches; not immediately critical
Next Steps: Specific actions to take
Common Issues
"Inventory file not found"
./scripts/collect-inventory.sh
"Claude command not found"
Use SDK mode instead:
export ANTHROPIC_API_KEY=sk-...
./bin/audit -claude=sdk
Or skip Claude analysis (findings only):
go run ./cmd/audit 2>/dev/null
"No findings found"
This is good! It means:
- Your packages are up-to-date
- Or they're not tracked in OSV.dev
- Check
audit_report.jsonfor details
Next Steps
- Review high-severity findings in
audit_report.html - Follow Claude's recommendations for each CVE
- Update packages in priority order (CRITICAL → HIGH → MEDIUM)
- Re-run audit after updates:
make audit-full
Output Files
| File | Purpose |
|---|---|
inventory.json |
System inventory (JSON) |
audit_report.json |
Findings + recommendations (JSON) |
audit_report.html |
Human-readable report (HTML) |
Advanced Usage
Filter by severity
# View only critical findings
jq '.findings[] | select(.cvss_score >= 9.0)' audit_report.json
Export for CI/CD
# Fail if critical vulnerabilities found
if [ $(jq '.summary.critical_findings' audit_report.json) -gt 0 ]; then
exit 1
fi
Schedule regular audits
# Add to crontab
0 9 * * 1 cd ~/Projects/package-review && make inventory && ./bin/audit && open audit_report.html
Documentation
For detailed information, see:
- README.md — Full documentation
- DATA_SOURCES_SPEC.md — Data source details
- Makefile — Available commands
Next: make audit-full then open audit_report.html