mandate-2-2-1-code-evaluation

name	mandate-2-2-1-code-evaluation
description	Evaluate compliance for Mandate 2.2.1 (AI Supply Chain Security) using repository code and configuration analysis. Use when producing deterministic pass/fail findings with severity, mapped mitigated vulnerabilities, and file-level evidence.

Mandate 2.2.1 Code Evaluation Skill

Mandate

• ID: 2.2.1
• Title: AI Supply Chain Security

Mitigates

• ML06 AI Supply Chain Attacks
• LLM03 Supply Chain
• ASI04 Agentic Supply Chain Vulnerabilities
• MCP06 Server Discovery and Verification

Inputs

• Repository root
• Build manifests and lockfiles
• CI/CD workflow files
• Dependency/security policy files

Workflow

1. Discover all dependency ecosystems in the repo.
   • Locate package*.json, pnpm-lock.yaml, requirements*.txt, poetry.lock, Pipfile.lock, go.mod, go.sum, Cargo.toml, Cargo.lock, pom.xml, build.gradle*, Gemfile.lock, .csproj, packages.lock.json.
2. Build a normalized dependency inventory.
   • Record package name, version constraint, source registry, environment (prod/dev), and lockfile status.
3. Detect pinning weaknesses.
   • Flag floating versions (*, latest, open ranges), mutable Git refs, and direct URL installs in production dependencies.
4. Verify inventory artifact generation.
   • Detect SBOM/AIBOM generation (CycloneDX/SPDX tools or scripts) in CI workflows or release scripts.
5. Verify automated vulnerability scans.
   • Detect scanners (osv-scanner, npm audit, pip-audit, trivy, grype, snyk, Dependabot, equivalent).
6. Verify policy enforcement.
   • Confirm CI fails builds on High/Critical findings; flag advisory-only scans.
7. Verify AI/MCP dependency tracking.
   • Confirm MCP servers/plugins/tools appear in inventory and scanning scope.
8. Emit findings with file/line evidence and concrete remediation.

Decision Rules

• Pass if all production dependencies are pinned/locked, SBOM/AIBOM is generated, vulnerability scans run automatically, and policy gates enforce failure thresholds.
• Fail if any required control is missing or weak.

Severity Rules

• Critical: no automated scanning and no lock/pin controls.
• High: scanning exists but not enforced, or multiple unpinned production dependencies.
• Medium: partial ecosystem coverage.

Output Template

{
  "mandate_id": "2.2.1",
  "status": "fail",
  "severity": "high",
  "vulnerability_tags": ["ML06", "LLM03", "ASI04", "MCP06"],
  "evidence": [
    {"file": ".github/workflows/ci.yml", "line": 38, "detail": "No dependency vulnerability scan step"},
    {"file": "package.json", "line": 19, "detail": "Production dependency uses floating version"}
  ],
  "remediation": "Pin production dependencies and enforce CI vulnerability scan gate."
}

Guardrails

• Use code and repo configuration evidence only.
• Always include file-level evidence for each finding.
• If required evidence is missing in code, mark control as not implemented.