| name | secure-sdlc-and-supply-chain |
| description | Applies NIST SSDF (SP 800-218) secure development and supply-chain controls including SBOM, pinning, secrets, SAST and MCP/A2A integrity. Use when hardening builds, dependencies, CI/CD, releases or agent tool supply chains. |
| aliases | ["secure-sdlc","software-supply-chain-security","ssdf-practice","nist-800-218","sbom","supply-chain","dependency-supply-chain","harden-dependencies","sbom-checks"] |
Secure SDLC and Supply Chain
When to use
Use this skill when establishing or changing how software is prepared, protected, produced and released: dependency management, build pipelines, artefact integrity, secret handling, static analysis, SBOM generation, or the supply chain of MCP servers, A2A agents and third-party tools that coding agents invoke.
When not to use
- Use
threat-modeling to identify STRIDE threats and trust boundaries before selecting controls.
- Use
guardrails-safety-patterns for runtime agent policy, input/output checks and tool allow-lists.
- Use
ci-cd-and-automation for pipeline structure, permissions and job design without a full SSDF/supply-chain focus.
- Use
release-engineering-and-progressive-delivery for canaries, rollback and progressive exposure once artefacts are already trustworthy.
Objective
Embed NIST SSDF (SP 800-218) practices into the development and delivery path so source, dependencies, builds, secrets, scans and agent-facing supply chains remain integrity-preserving, auditable and least-privilege by default.
Procedure
- Map the change to SSDF practice groups: prepare the organisation (standards, approved sources), protect software (source and build integrity), produce well-secured software (secure coding and review), and respond to vulnerabilities (fix and regression).
- Inventory third-party and first-party dependencies, actions, base images, MCP servers, A2A peers and plugins. Prefer approved registries and signed or verified sources where the project supports them.
- Pin versions for packages, container images and GitHub Actions (or equivalent). Record justification when floating tags are unavoidable and schedule follow-up.
- Keep secrets out of source, logs and model context. Use secret stores, short-lived credentials and rotation paths; fail closed when secrets are missing rather than embedding defaults.
- Enable or preserve SAST, dependency scanning and secret scanning in CI. Treat findings as release-relevant; do not disable gates to force a green build.
- Produce or update an SBOM for released artefacts and retain it with the build provenance needed for incident response.
- For MCP/A2A supply chain: pin server versions or digests, restrict tool permissions, validate schemas, log peer identity and refuse unreviewed remote tools that can execute privileged actions.
- Document how vulnerabilities are triaged, fixed and regression-tested, including who owns response for agent-tool dependencies.
Required outputs
# Secure SDLC / supply-chain checklist: <change>
## SSDF mapping
- Prepare:
- Protect:
- Produce:
- Respond:
## Dependencies and pins
| Component | Version / digest | Source | Notes |
## Secrets and credentials
- Storage:
- Rotation:
- Log / context redaction:
## Scanning and gates
- SAST:
- Dependency / secret scan:
- Release blockers:
## SBOM and provenance
- SBOM location:
- Build identity / attestation (if any):
## MCP / A2A supply chain
- Peers and versions:
- Permission scope:
- Validation and audit:
## Residual risks
Rules
- Do not commit secrets, tokens or private keys.
- Do not unpin or broaden dependencies without recording risk and owner.
- Do not bypass SAST, secret scan or dependency gates without explicit human approval.
- Do not treat an SBOM as optional documentation; keep it with release artefacts when releasing.
- Do not trust MCP/A2A peers solely because they are configured in a client; validate identity, version and permissions.
- Do not claim full NIST certification; align practices to SP 800-218 intent and project policy.
Related skills
ci-cd-and-automation — safe pipeline design and least-privilege jobs
guardrails-safety-patterns — runtime agent and application controls
threat-modeling — identify threats that supply-chain controls must address
release-engineering-and-progressive-delivery — safe rollout after integrity gates pass
References
Verification