agentleFS
Sign inStart free
TheCollieryTheColliery/CoalMine/skills/supply-chain-audit/SKILL.md

supply-chain-audit

Skill15 starsChanged 6 days agoApache-2.0View on GitHub

Software supply chain audit — dependencies (CVEs, maintenance, licenses, transitive risk), build/CI integrity (SHA-pinned actions, lockfile, CI-only release), artifact integrity (checksums, signing, SBOM). Triggers on: "/supply-chain-audit", "supply-chain-audit", "dependency audit". Run before adding a dep, before a release, or for periodic review. Reports; does not change deps unless asked.

9 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Don't auto-change deps — report + recommend.

    line 38
  2. Very specific:

    After the report, present via ask_question:

    line 42
  3. Distinctive:

    NEVER auto-fix: dep version bump, lockfile regen (re-resolves entire transitive tree).

    line 47

“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Supply-Chain Audit

<!-- SHARED:LANGUAGE_HEADER -->

Audit what the project trusts: deps, build pipeline, shipped artifact. Report; do NOT change deps unless asked.

1. Dependencies

  • Scope — honor .coalmine.json packageManifests if set: scan exactly those manifest/lockfile paths; else infer by inspecting the repo.
  • CVEs — run ecosystem auditor; cross-check every hit in GHSA/OSV/NVD. Cite advisory ID + affected range + fixed version. (Invoke source-grounding — never from memory.)
  • Maintenance — last release, commit recency, bus-factor, archived/deprecated flag.
  • License — flag copyleft inside permissive project, missing/unknown license.
  • Transitive — full tree; name the parent to bump for a transitive fix.
  • Behavior — phone home? install scripts? unexpected egress?

2. Build / CI

  • CI-only release builds?
  • Actions pinned to commit SHA (not floating tag)?
  • Lockfile committed + enforced in CI?
  • Minimal token scope? No pull_request_target?

3. Artifact

  • SHA-256 checksums published for every binary?
  • Signed (Authenticode/GPG)? Gap documented honestly?
  • SBOM generated?
  • User can verify before running?

Tooling

Per-ecosystem vuln/license/outdated commands + offline fallback: read references/tooling.md when selecting scanners.

Discipline

  • Ground every CVE/fixed-version in an advisory. Never from memory.
  • Don't auto-change deps — report + recommend; user decides (bumps break builds).
  • State what was NOT scanned. Blocked network scans → lockfile inspection fallback (see references/tooling.md), mark live checks N-A.

Fix mode (choice-gated)

After the report, present via ask_question:

  • Pin safe now — commit already-present unchanged lockfile, pin CI action to current SHA, add missing checksum step. Each: checkpoint → apply → verify.
  • Let me pick — user-selected fixes only.
  • Report only — change nothing.

NEVER auto-fix: dep version bump, lockfile regen (re-resolves entire transitive tree).

Grants & denials (CLASSIFY-BLOCK)

classstep it powersgranton denial
readscan manifests/lockfiles; run the ecosystem auditorRead·Grep·Glob·Bash (read-only)refuse that manifest, name it in "Not scanned" — never a clean bill
writeFix mode's pin/commit, incl. checkpoint → apply → verifyEdit·Write·Bash (checkpoint/verify need exec)report the pin as NOT applied AND the checkpoint/verify as NOT available, never claim done
networkGHSA/OSV/NVD cross-checkWebSearch·WebFetch (or delegated to source-grounding)⚠️ unverified: check [advisory ID]
<!-- SHARED:CLASSIFY_BLOCK -->

Output

| package | direct/transitive | issue | severity | advisory | fixed-in | action | Build+artifact checklist · Summary (counts + top fixes) · Not scanned

For ReportFindings: file = the manifest/lockfile path that named the package (packageManifests, or the inferred one); line is best-effort (the pin/version line if easily found) or omitted and named imprecise per the shared reporting rail — never fabricate a line.

<!-- SHARED:REPORTING_FOOTER --><!-- SHARED:ORCHESTRATION --><!-- SHARED:ESCALATION_FOOTER -->

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free