feat: add plugin marketplace and skills install#1
Merged
Conversation
* Add .claude-plugin/marketplace.json and plugin.json so the skill is installable via `/plugin marketplace add petems/unsloppify` and `/plugin install unsloppify@unsloppify` * Surface `npx skills add petems/unsloppify` as a second install path (works with Claude Code, Codex, Cursor) via the existing skills/unsloppify/ layout * Add tests/test_skill_metadata.py validating SKILL.md frontmatter (allowed-keys set, required name + description), marketplace + plugin manifest schema, name cross-check, and plugin.json/pyproject.toml version sync * Add .github/workflows/skill-validate.yml with three jobs: pytest metadata tests, npx skills add codex-target smoke install, and the canonical `claude plugin validate` (continue-on-error) * Update README install section to surface both install paths * Include .claude-plugin/ in the sdist build Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
📝 WalkthroughWalkthroughThe PR adds Claude Code plugin manifests and a marketplace entry for "unsloppify", introduces pytest metadata validation for SKILL.md and plugin manifests with cross-file checks (name and version sync, directory matching), adds a GitHub Actions workflow to run tests, smoke-install, and plugin validation, and updates README and sdist packaging to include plugin artifacts. ChangesClaude Code Plugin Setup and Validation
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.github/workflows/skill-validate.yml:
- Around line 64-77: Remove the permissive CI flag so validator failures fail
the job: delete or set continue-on-error to false for the job that runs the
Claude Code plugin validator (the job containing the "Run the Claude Code plugin
validator" step which runs `npx --yes `@anthropic-ai/claude-code` plugin validate
.`), ensuring the workflow's continue-on-error: true entry is removed/disabled
so manifest/validator regressions block merges.
In `@tests/test_skill_metadata.py`:
- Around line 105-109: The test currently only checks that "source" exists and
conditionally validates its string form; change it to require that
plugins[].source is a string unconditionally by asserting
isinstance(entry["source"], str) (e.g., in tests/test_skill_metadata.py where
entry is used) and then keep the startswith("./") assertion for local paths —
this ensures non-string sources fail fast before calling
entry["source"].startswith.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 4d17786d-cd52-4c36-be9d-5ad8e4f0d93c
📒 Files selected for processing (6)
.claude-plugin/marketplace.json.claude-plugin/plugin.json.github/workflows/skill-validate.ymlREADME.mdpyproject.tomltests/test_skill_metadata.py
* Replace --all with --global --skill '*' so --agent codex is honored * Verify ~/.agents/skills/unsloppify/SKILL.md (codex is a universal agent) * Use skills list --global to surface globally-installed skill Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
♻️ Duplicate comments (1)
.github/workflows/skill-validate.yml (1)
65-77:⚠️ Potential issue | 🟠 Major | ⚡ Quick winMake the canonical plugin validator blocking.
Line 65 (
continue-on-error: true) allows validator regressions to merge silently, and the explanatory note on Lines 75-77 bakes that behavior in.Suggested fix
claude-plugin-validate: name: claude plugin validate (canonical validator) runs-on: ubuntu-latest - continue-on-error: true steps: - uses: actions/checkout@v6 - uses: actions/setup-node@v5 with: node-version: "22" - name: Run the Claude Code plugin validator run: | set -euxo pipefail # Validates marketplace.json, plugin.json, and skill frontmatter using - # the canonical CLI. Job is marked continue-on-error in case the npm - # package name shifts; the metadata-tests job is the load-bearing - # check. + # the canonical CLI. npx --yes `@anthropic-ai/claude-code` plugin validate .🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/workflows/skill-validate.yml around lines 65 - 77, The workflow currently silences failures by setting continue-on-error: true for the "Run the Claude Code plugin validator" job; change that so the canonical plugin validator is blocking by removing or setting continue-on-error to false for that job (the step named "Run the Claude Code plugin validator") and ensure the explanatory comment about marking the job non-blocking is updated/removed so regressions will fail the workflow.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Duplicate comments:
In @.github/workflows/skill-validate.yml:
- Around line 65-77: The workflow currently silences failures by setting
continue-on-error: true for the "Run the Claude Code plugin validator" job;
change that so the canonical plugin validator is blocking by removing or setting
continue-on-error to false for that job (the step named "Run the Claude Code
plugin validator") and ensure the explanatory comment about marking the job
non-blocking is updated/removed so regressions will fail the workflow.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 190081b8-10a2-4e2e-ab67-edd63ea5067d
📒 Files selected for processing (1)
.github/workflows/skill-validate.yml
* skill-validate.yml: drop continue-on-error on canonical validator so manifest regressions fail the job instead of silently passing * test_skill_metadata.py: require plugins[].source to be a non-empty string before checking the ./ prefix Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
npx skills addCLI — instead of the manual copy-into-~/.claude/skills/workflow the README previously documented.skill-validate.ymlworkflow so future edits toSKILL.md,marketplace.json, orplugin.jsoncan't silently break installation.What changed
.claude-plugin/marketplace.json+.claude-plugin/plugin.jsonat the repo root. Plugin source is"./"so the whole repo is the plugin and the existingskills/unsloppify/is auto-discovered.tests/test_skill_metadata.py(8 tests): SKILL.md frontmatter parse + allowed-keys check (name,description,license,allowed-tools,metadata— anything else makes the Claude app reject the skill), marketplace and plugin schema, name cross-check,plugin.json↔pyproject.tomlversion sync, skill-dir/name match..github/workflows/skill-validate.yml: three jobs —metadata-tests(pytest),skill-install-smoke(npx skills add ./ --agent codexinto an isolated$HOME, verify SKILL.md lands at~/.codex/skills/unsloppify/SKILL.md), andclaude-plugin-validate(npx @anthropic-ai/claude-code plugin validate .,continue-on-erroruntil the npm package name is confirmed)./plugin marketplace add petems/unsloppifyandnpx skills add petems/unsloppify..claude-plugin/added to the sdist include list inpyproject.toml.Why codex for the smoke test
Claude Code users will install via the plugin marketplace path (covered by
claude-plugin-validate). Targetingcodexwithnpx skills addbroadens coverage by proving the skill's frontmatter + layout works with non-Claude agents too.Test plan
pytest tests/test_skill_metadata.py -v— 8 passed locallypytest) — 40 passed locallymypy src tests/test_skill_metadata.pycleanruff check tests/test_skill_metadata.pycleanskill-validate.yml / metadata-testsgreen on PRskill-validate.yml / skill-install-smokegreen — confirms the codex install path lands at~/.codex/skills/unsloppify/SKILL.md(if the path differs the assertion will fail loudly and we adjust)skill-validate.yml / claude-plugin-validate(informational,continue-on-error)ci.ymlstill green/plugin marketplace add petems/unsloppifyfollowed by/plugin install unsloppify@unsloppifysucceeds and/unsloppify:unsloppifyis invocable🤖 Generated with Claude Code
Summary by CodeRabbit
New Features
Documentation
Tests
Chores