Github Supply Chain

GitHub Supply Chain Audit

Audit the external code and identity surface that GitHub Actions exposes: third-party actions, dependencies, build outputs, and the credentials workflows use to reach cloud providers. This is a defensive (find & fix) skill — find weaknesses and produce a remediation report.

When this skill applies

• Reviewing .github/workflows/*.yml for hardening gaps
• Evaluating dependency hygiene (Dependabot config, Renovate, lockfiles, audit policies)
• Checking SBOM presence and quality
• Auditing how workflows authenticate to AWS/GCP/Azure (long-lived keys vs OIDC)
• Triage after a public Actions compromise (e.g., tj-actions/changed-files class of incidents)

Use a different skill for: repository governance like branch protection (see github-repo-hardening), application-code vulnerabilities (see saas-code-security-review).

Workflow

Follow the 5-phase audit workflow defined in ../_shared/audit-workflow.md. Skill-specific guidance below.

Phase 1: Scope confirmation

Ask the user (or confirm from context):

• Single repo or org-wide?
• Public, internal, or private repo?
• Has GitHub Advanced Security (GHAS) license? (affects which features are available)
• Are workflows reaching cloud providers? Which?

Phase 2: Inventory

Collect:

# List all workflow files
find .github/workflows -name '*.yml' -o -name '*.yaml'

# Extract every external action reference
grep -rEho 'uses:\s*[^@\s]+@[^\s]+' .github/workflows/ | sort -u

# List dependency manifests
find . -name 'package.json' -o -name 'package-lock.json' -o -name 'pnpm-lock.yaml' \
       -o -name 'requirements*.txt' -o -name 'Pipfile.lock' -o -name 'poetry.lock' \
       -o -name 'go.mod' -o -name 'go.sum' -o -name 'Gemfile.lock' -o -name 'Cargo.lock' \
       -o -name 'pom.xml' -o -name 'build.gradle*' | grep -v node_modules

Also check:

• .github/dependabot.yml presence and ecosystems covered
• renovate.json / renovate.json5 if Renovate is used instead
• .github/workflows/codeql.yml for code scanning
• Repository settings for "Dependency graph" and "Dependabot alerts" (via API or gh api)

Phase 3: Detection — the checks

Apply every check below. Reference the linked file when the check is non-trivial.

Action pinning and provenance — see references/actions-hardening.md

• GHSC-PIN-1 Every third-party action pinned by full commit SHA, not by tag or branch. Tags are mutable.
• GHSC-PIN-2 First-party actions/* may be pinned by major version (@v4) since GitHub owns them, but SHA is still preferred for high-security contexts.
• GHSC-PIN-3 Comment next to each pin showing the human-readable version: uses: foo/bar@ # v1.2.3. Without it, future updates are blind.
• GHSC-PIN-4 No use of @main, @master, @latest, or floating tags anywhere.

Workflow permissions — see references/actions-hardening.md

• GHSC-PERM-1 permissions: {} declared at the workflow OR job level. Defaulting to permissive grants GITHUB_TOKEN write access to most scopes.
• GHSC-PERM-2 When write permissions are needed, scoped narrowly (e.g., contents: read, pull-requests: write) not write-all.
• GHSC-PERM-3 pull_request_target triggers reviewed carefully — they run with secrets and write tokens against PRs from forks. Cross-reference with any checkout of PR head SHA.

Secrets handling

• GHSC-SEC-1 No secrets in plaintext in workflow files, even masked.
• GHSC-SEC-2 Secrets not passed to third-party actions unless that action's source is reviewed and pinned.
• GHSC-SEC-3 Environment secrets used for production (with required reviewers) rather than repo-level secrets.
• GHSC-SEC-4 No echo "$SECRET" patterns — GitHub masks known secrets in logs but transformations (base64, JSON-wrap) break the mask.

Dependency hygiene — see references/dependency-review.md

• GHSC-DEP-1 Dependabot or Renovate configured for every ecosystem present.
• GHSC-DEP-2 Lockfile present for every package manager that supports one (package-lock.json, pnpm-lock.yaml, yarn.lock, poetry.lock, Pipfile.lock, go.sum, Cargo.lock).
• GHSC-DEP-3 dependency-review-action runs on every PR to flag new vulnerable deps before merge.
• GHSC-DEP-4 Internal packages scoped (@yourorg/) on a private registry, with namespace held on the public registry to prevent dependency confusion.
• GHSC-DEP-5 Install commands use --ignore-scripts where feasible, OR malicious-package detection runs before install (Socket, Snyk, etc.).

SBOM and provenance — see references/sbom-generation.md

• GHSC-SBOM-1 SBOM generated for every release artifact (CycloneDX or SPDX format).
• GHSC-SBOM-2 SBOM stored as a release asset or in a queryable store, not just printed to logs.
• GHSC-SBOM-3 For container images, SBOM attached as an OCI attestation (Cosign + Sigstore).
• GHSC-SBOM-4 Build provenance (SLSA L2+) attested for production artifacts.

OIDC and cloud authentication — see references/oidc-cloud-auth.md

• GHSC-OIDC-1 Long-lived cloud credentials (AWS access keys, GCP service account keys, Azure SP secrets) replaced by OIDC federation.
• GHSC-OIDC-2 OIDC trust policy scoped to specific repo, ref, and environment claims — not just the org.
• GHSC-OIDC-3 Audience claim set explicitly (default sts.amazonaws.com for AWS) rather than allowing any.
• GHSC-OIDC-4 Wildcards in sub claim limited (e.g., repo:org/repo:ref:refs/heads/main not repo:org/*).

Build environment

• GHSC-ENV-1 Self-hosted runners isolated (ephemeral, single-job, not on persistent VMs with secrets).
• GHSC-ENV-2 Self-hosted runners not enabled on public repos without isolation.
• GHSC-ENV-3 Workflow defaults block does not silently shell out to bash -e -o pipefail with unvalidated inputs.

Phase 4: Triage

Critical class examples for this skill:

• Third-party action with write token access pinned by mutable tag (one upstream takeover = full repo compromise)
• Long-lived AWS access key with admin policy in repo secrets
• pull_request_target checking out PR head SHA then running PR code with secrets

Phase 5: Report

Use ../_shared/findings-schema.md. Prefix IDs with GHSC-.

Outputs

The skill produces:

1. A Markdown report at audits/github-supply-chain//.md
2. (Optional, on request) A hardened workflow template based on assets/hardened-workflow.yml
3. (Optional, on request) A migration plan from long-lived keys to OIDC

References

• references/actions-hardening.md — Pinning, permissions, pull_request_target patterns
• references/dependency-review.md — Dependabot vs Renovate, dependency-review-action, lockfile policy
• references/sbom-generation.md — CycloneDX/SPDX, Cosign attestations, SLSA levels
• references/oidc-cloud-auth.md — Trust policy templates for AWS, GCP, Azure
• assets/hardened-workflow.yml — Production-grade workflow template

Scripts

• scripts/extract_action_pins.sh — Pulls every uses: reference and flags non-SHA pins
• scripts/check_lockfiles.sh — Verifies presence of lockfile for each detected ecosystem

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: hlsitechio
• Source: hlsitechio/claude-skills-security
• License: MIT

Install and usage instructions live in the source repository linked above.