agentleFS
odere-proodere-pro/claude-software-3-0-marketplace/.github/CLAUDE.md

claude-software-3-0-marketplace

CLAUDE.md2 starsChanged 5 months agoMITView on GitHub

Author-only. Continuous integration and the supply-chain hardening for the registry. - ci.yml — always-on (every push + PR). Installs jq + shellcheck and runs bash tests/gates/run-all.sh. This is the gate that must stay green. The checkout uses fetch-depth: 0 (full history) so the commit-author backstop gate 17-commit-author.sh can scan every commit (roadmap P14/D15). It also carries two PR-only jobs: - history-secret-scan — runs the license-free gitleaks CLI binary (pinned version, SHA-256 verified before use) over the full git history to…

  • Reads credentials

The file

Files
CLAUDE.md

.github/ — CI, supply chain, governance wiring

Author-only. Continuous integration and the supply-chain hardening for the registry.

Workflows (workflows/)

  • ci.yml — always-on (every push + PR). Installs jq + shellcheck and runs bash tests/gates/run-all.sh. This is the gate that must stay green. The checkout uses fetch-depth: 0 (full history) so the commit-author backstop gate 17-commit-author.sh can scan every commit (roadmap P14/D15). It also carries two PR-only jobs:
    • history-secret-scan — runs the license-free gitleaks CLI binary (pinned version, SHA-256 verified before use) over the full git history to FAIL the PR on any committed secret (roadmap P-history-secret-scan; active now, the repo is public, D16). It shares the same fetch-depth: 0 checkout pattern. The CLI is used instead of gitleaks/gitleaks-action, which requires a paid GITLEAKS_LICENSE on org-owned repos. It runs only on pull_request, so it never blocks a push.
    • reproducible-diff — see below.
  • scorecard.yml — OpenSSF Scorecard. Active (the repo is public, D16): weekly schedule, push to main, branch_protection_rule, and manual workflow_dispatch. Publishes the score to the public Scorecard API and uploads SARIF to code-scanning.
  • codeql.yml — CodeQL SAST over languages: actions (scans the workflow YAML itself; there is no compiled language here). Active now the repo is public (D16): push/PR to main, a weekly schedule, and manual workflow_dispatch.
  • audit.yml — the consolidated cross-repo provenance audit (roadmap P4, Phase 3). Scheduled (weekly cron) + workflow_dispatch. It re-derives, live, for every plugin listed in marketplace.json: the repo exists and its plugin.json is fetchable/valid, source.repo matches odere-pro/<repo>, and the listed repo ships no marketplace.json of its own (it does not compare entry.name to plugin.json.name — the entry name may differ via --name, D13). It reuses the vet-candidate.sh probes (DRY — no fork) via scripts/audit-cross-repo.sh. It is advisory / notify-only (D3): it never hard-FAILs, SKIPs on any non-200/network/gh-auth error (a connectivity control probe distinguishes a transient outage from real drift). Per-entry checks are a no-op while the registry is empty (N=0); the About/topics check runs unconditionally after the connectivity control probe passes. It is deliberately not a tests/gates/ push-suite gate — it makes network calls, which never enter run-all.sh (D3/D4). Auth is the default GITHUB_TOKEN exported as GH_TOKEN (reads public plugin repos, no extra secret, D17/Q1). Two additional checks extend the per-entry loop and post-loop (Phase 3):
    • CHECK A — Install-coordinate validity (DRIFT-level): MISSING_NAME is added to the per-entry relevance filter in audit-cross-repo.sh. A blank plugin.json.name breaks <plugin>@odere-pro install resolution (provenance drift). Source-fetchable is already covered by PLUGIN_JSON_UNREADABLE — no new probe needed (ADR: MISSING_NAME crosses the D13 line for coordinate-validity but stays excluded as an add-time quality floor at vet-time).
    • CHECK B — About/topics advisory (ADVISORY-level, registry repo only): checks the odere-pro/claude-software-3-0-marketplace repo for recommended discovery topics (claude-code, marketplace, claude-code-plugins) and a non-empty About description. Findings go to "advisories" in the report — never counted as drift, never escalate the drift issue. Two notify channels: on drift (or drift + advisories) → opens/updates the provenance-drift issue ("cross-repo audit: provenance drift detected", advisories appended in its body). On advisory-only (drift empty) → opens/updates a separate advisory issue ("cross-repo audit: discoverability advisory") so advisory notices never co-mingle with the provenance-drift issue. The open-or-update idiom is identical across both channels. Permissions are least-privilege: contents: read to read the checked-out manifest/scripts plus issues: write for the notify — both narrow scopes G16 permits without an allowlist entry (issues: write, like security-events/id-token write, is not the high-blast-radius contents: write). The single actions/checkout is SHA-pinned (G15).

The ci.yml reproducible-diff job (PR-only) runs scripts/reproducible-diff.sh (roadmap P15): a network-free "green ⇒ machine-generated" check that the manifest's structural fields (source/license/homepage, in canonical jq --indent 2 form — the same formatter .claude/hooks/json-format.sh uses) reproduce the script-emitted shape. It excludes the LLM-curated description/keywords and is not auto-merge — CODEOWNERS still governs merge (D7). It lives under scripts/ (not tests/gates/) so it never enters the always-on push suite run-all.sh.

The audit.yml job runs scripts/audit-cross-repo.sh (roadmap P4): the scheduled, advisory provenance re-derivation described above. Like reproducible-diff.sh it lives under scripts/ (not tests/gates/) so it never enters the push suite — it makes network calls and is notify-only (D3). It drives vet-candidate.sh per listed entry rather than re-implementing the probes (DRY).

Supply chain

  • All actions are pinned by full commit SHA with a trailing # vX.Y.Z comment — an OpenSSF "Pinned-Dependencies" signal.
  • dependabot.yml watches the github-actions ecosystem weekly and bumps the SHA pins (it updates the pin and the comment together), committing with a ci: prefix.

Governance

  • CODEOWNERS — * @odere-pro; enables code-owner review on branch-protected PRs (a Scorecard "Code-Review" signal).

No release.yml / tag-release flow: a registry has no versioned artifact of its own (each listed plugin owns its version). Revisit only if the registry itself ever becomes versioned.

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.