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
- 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.
- Build a normalized dependency inventory.
- Record package name, version constraint, source registry, environment (
prod/dev), and lockfile status.
- Detect pinning weaknesses.
- Flag floating versions (
*, latest, open ranges), mutable Git refs, and direct URL installs in production dependencies.
- Verify inventory artifact generation.
- Detect SBOM/AIBOM generation (CycloneDX/SPDX tools or scripts) in CI workflows or release scripts.
- Verify automated vulnerability scans.
- Detect scanners (
osv-scanner, npm audit, pip-audit, trivy, grype, snyk, Dependabot, equivalent).
- Verify policy enforcement.
- Confirm CI fails builds on High/Critical findings; flag advisory-only scans.
- Verify AI/MCP dependency tracking.
- Confirm MCP servers/plugins/tools appear in inventory and scanning scope.
- 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.
1---2name: mandate-2-2-1-code-evaluation3description: 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.4---56# Mandate 2.2.1 Code Evaluation Skill78## Mandate9- ID: 2.2.110- Title: AI Supply Chain Security1112## Mitigates13- ML06 AI Supply Chain Attacks14- LLM03 Supply Chain15- ASI04 Agentic Supply Chain Vulnerabilities16- MCP06 Server Discovery and Verification1718## Inputs19- Repository root20- Build manifests and lockfiles21- CI/CD workflow files22- Dependency/security policy files2324## Workflow251. Discover all dependency ecosystems in the repo.26 - 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`.272. Build a normalized dependency inventory.28 - Record package name, version constraint, source registry, environment (`prod`/`dev`), and lockfile status.293. Detect pinning weaknesses.30 - Flag floating versions (`*`, `latest`, open ranges), mutable Git refs, and direct URL installs in production dependencies.314. Verify inventory artifact generation.32 - Detect SBOM/AIBOM generation (CycloneDX/SPDX tools or scripts) in CI workflows or release scripts.335. Verify automated vulnerability scans.34 - Detect scanners (`osv-scanner`, `npm audit`, `pip-audit`, `trivy`, `grype`, `snyk`, Dependabot, equivalent).356. Verify policy enforcement.36 - Confirm CI fails builds on High/Critical findings; flag advisory-only scans.377. Verify AI/MCP dependency tracking.38 - Confirm MCP servers/plugins/tools appear in inventory and scanning scope.398. Emit findings with file/line evidence and concrete remediation.4041## Decision Rules42- Pass if all production dependencies are pinned/locked, SBOM/AIBOM is generated, vulnerability scans run automatically, and policy gates enforce failure thresholds.43- Fail if any required control is missing or weak.4445## Severity Rules46- Critical: no automated scanning and no lock/pin controls.47- High: scanning exists but not enforced, or multiple unpinned production dependencies.48- Medium: partial ecosystem coverage.4950## Output Template51```json52{53 "mandate_id": "2.2.1",54 "status": "fail",55 "severity": "high",56 "vulnerability_tags": ["ML06", "LLM03", "ASI04", "MCP06"],57 "evidence": [58 {"file": ".github/workflows/ci.yml", "line": 38, "detail": "No dependency vulnerability scan step"},59 {"file": "package.json", "line": 19, "detail": "Production dependency uses floating version"}60 ],61 "remediation": "Pin production dependencies and enforce CI vulnerability scan gate."62}63```6465## Guardrails66- Use code and repo configuration evidence only.67- Always include file-level evidence for each finding.68- If required evidence is missing in code, mark control as not implemented.
jefflyt