Security Analyzer

Analyze environments for vulnerabilities, fetch current CVE/exploit data, and generate phased remediation plans with TDD validation.

Quick Start

When the user requests a security scan:

1. Run environment discovery: python .claude/skills/security-analyzer/scripts/discover_env.py .
2. Save output to inventory.json
3. Run vulnerability scan: python .claude/skills/security-analyzer/scripts/fetch_vulns.py inventory.json
4. Save output to scan_results.json
5. Generate reports: python .claude/skills/security-analyzer/scripts/generate_report.py scan_results.json inventory.json

Workflow

Phase 1: Environment Discovery

Scan working directory for:

• Dependencies: package.json, requirements.txt, Gemfile, go.mod, Cargo.toml, pom.xml
• Containers: Dockerfile, docker-compose.yml, kubernetes/*.yaml
• Cloud IaC: terraform/*.tf, cloudformation/*.yaml, *.bicep
• Secrets: .env* files (flag exposure risk, never log values)

Run the discovery script:

python .claude/skills/security-analyzer/scripts/discover_env.py /path/to/project > inventory.json

Phase 2: Vulnerability Intelligence

Fetch current threat data using the vulnerability scanner:

python .claude/skills/security-analyzer/scripts/fetch_vulns.py inventory.json > scan_results.json

For CISA KEV and additional context, supplement with:

WebSearch: "CVE-XXXX-YYYY CISA KEV exploit"

Phase 3: Risk Scoring

The scanner calculates risk scores using:

Risk = (CVSS * 0.3) + (Exploitability * 0.3) + (Criticality * 0.2) + (Exposure * 0.2)

Exploitability: 10=CISA KEV, 7=public exploit, 3=theoretical
Criticality: 10=auth/payment, 5=core business, 2=logging
Exposure: 10=internet-facing, 5=internal, 2=air-gapped

Phase 4: Phased Remediation

Generate reports with fix commands and validation tests:

python .claude/skills/security-analyzer/scripts/generate_report.py scan_results.json inventory.json

Each finding includes:

1. Vulnerability details + risk score
2. Actual fix code/patch (not just recommendations)
3. Pre-fix test (proves vuln exists)
4. Remediation unit tests (tests the fix code)
5. Post-fix validation (proves vuln resolved)

Phase 5: Reports

Output two reports:

• security-report-technical.md — Full details for engineers
• security-report-executive.md — Summary for leadership

See references/report-templates.md for output structure.

TDD Pattern

For each vulnerability, generate three test types:

def test_vuln_exists():
    """PASS before fix, FAIL after"""
    assert is_vulnerable("component") == True

def test_fix_works():
    """Unit test for remediation code"""
    result = apply_fix(vulnerable_config)
    assert result.is_secure()

def test_vuln_resolved():
    """FAIL before fix, PASS after"""
    assert is_vulnerable("component") == False

Fix Types by Finding

Example Interaction

User: “Run a security scan on this project”

Claude:

1. Discovers 47 npm dependencies, 3 Dockerfiles, 2 Terraform configs
2. Fetches current CVE data from OSV.dev
3. Identifies 12 vulnerabilities (2 critical, 4 high, 6 medium)
4. Generates phased remediation plan with:
   • Actual fix commands (npm install lodash@4.17.21)
   • Code patches for IaC misconfigurations
   • TDD tests proving each fix works
5. Outputs technical and executive reports