Claude Code Plugin Marketplace Template
A ready-to-use GitHub template for building and distributing Claude Code plugin marketplaces, with scaffolding, validation, and CI/CD built in.
Skill declares least-privilege allowed-tools (Read/Grep/Glob), defaults to read-only with user confirmation before edits, includes red-flag stop lists and an escalation/audit path; attribution and data-flow disclosure are largely complete. Deducted for: unverified publisher and reference docs (hooks-schema.md) that demonstrate persisting API keys into CLAUDE_ENV_FILE — not executed here but a sensitive-credential-on-disk pattern left without security caveats.
Documentation is internally consistent: checklists, common-mistakes and a diagnostic table cross-validate each other, with explanations for silent failures, timeouts and naming conflicts. Deducted because static review cannot execute key paths (/plugin-development:validate etc.), the repo CI only validates structure rather than the skill's core flows, and no committed test suite exists — capped below 10 by static calibration.
The description precisely enumerates trigger directories/files (.claude-plugin/, plugin., etc.) with clear fit, workflows and troubleshooting; semantic trigger precision is good. Deducted because functionality depends entirely on Claude Code plus overseas docs (docs.claude.com/code.claude.com) with no Chinese-language support or mainland-China reachability notes.
Progressive disclosure via linked schemas/templates/examples/best-practices, thorough naming and organization conventions, SemVer/CHANGELOG guidance, and clear MIT license. Deducted because the skill itself lacks an independent version/changelog, the LICENSE copyright holder is the 'Your Organization' template placeholder, and maintenance/update responsibility rests on an unverified GitHub account.
The skill substitutes vetted templates, checklists and reusable commands for freeform generation, plausibly saving silent-failure debugging cost, with directly followable workflows. Deducted because static review cannot confirm outputs are directly usable; the hello-world example and CI badge are not third-party execution evidence — capped below 7 by calibration.
Primary auditable material exists: CI workflow, schema docs, examples, and cross-references to official Anthropic documentation, with clean fact/inference separation. Deducted because there is no independent reproduction and no committed skill-level test results; evidence is limited to the repository's own files — capped at 5 by static calibration.
- This is a static source review; no commands or validation flows were executed and real-world usability is not independently reproduced.
- The CLAUDE_ENV_FILE example in hooks-schema.md demonstrates persisting API keys into an environment file; users copying this pattern should assess credential-at-rest risk themselves.
- The skill depends entirely on the Claude Code platform and overseas official docs (docs.claude.com / code.claude.com); reachability from mainland-China networks is not guaranteed and no Chinese-language support is provided.
- Publisher is an unverified GitHub account and the LICENSE copyright holder is a template placeholder; no committed maintenance or update path.
- The skill has no version number or changelog; users must diff repository history when upgrading.
What does this skill do, and when should you use it?
This is a GitHub template repository for creating and distributing Claude Code plugin marketplaces. It ships with a schema-compliant marketplace., a plugin-development toolkit offering 7 slash commands, a plugin-authoring skill, a plugin-reviewer agent, and GitHub Actions workflows that validate every plugin on each push and PR. A hello-world example plugin and local docs covering commands, hooks, skills, and sub-agents are included. It targets engineering teams that need a standardized way to distribute Claude Code plugins to their team or community.
Provides a cloneable marketplace repository structure (.claude-plugin/marketplace.); scaffolds plugin skeletons and adds commands, skills, agents, and hooks via commands like /plugin-development:init; runs /plugin-development:validate for schema and structure checks; creates a local dev marketplace via /plugin-development:test-local; runs automated plugin validation through GitHub Actions on every push and PR; and exposes a plugin-authoring skill that gives read-only guidance, checklists, and common-mistake warnings whenever context includes .claude-plugin/, plugin., hooks/, and related paths.
- A platform team establishing a single Claude Code plugin distribution channel for the whole organization, using .claude/settings. so members get the marketplace automatically on cloning.
- A plugin author starting from scratch, using init/add-command/add-skill to generate standard structure instead of hand-writing plugin..
- An open-source maintainer who wants every PR automatically validated, baking /plugin-development:validate checks into CI.
- A developer new to component directory conventions (components go at plugin root, not inside .claude-plugin/) relying on the skill's red-flag list to avoid silent failures.
- A team needing a pre-release deep audit, delegating to the plugin-reviewer agent for release-readiness review.
What are this skill's strengths and limitations?
- Complete workflow from template to example plugin (hello-world), eliminating marketplace boilerplate.
- Built-in validate command and CI workflows; the docs claim validation catches about 90% of issues before debugging.
- The skill encodes many real failure modes (reserved words, ${CLAUDE_PLUGIN_ROOT} paths, chmod +x) and explicitly flags scenarios that fail silently.
- Local documentation covers plugins, hooks, settings, commands, skills, and sub-agents.
- Deeply tied to Claude Code's plugin and slash-command system; not usable outside Claude Code as-is.
- No standalone automated test suite is documented; validation relies on the validate command and CI checks.
- No migration guidance for platforms beyond Claude Code (other coding agents, raw APIs).
- The repo presents no adoption metrics (stars, downloads) or production case studies, so maturity must be assessed independently.
How do you install this skill?
- Click "Use this template" on GitHub, then clone your new repository. 2. Edit .claude-plugin/marketplace. with your organization details. 3. Start Claude Code and run: /plugin marketplace add . then /plugin install plugin-development@my-team-marketplace. 4. To install published plugins from GitHub: /plugin marketplace add your-org/your-marketplace-name, then /plugin install plugin-name@your-marketplace-name. Note: system requirements (e.g., runtime versions) are not documented.
How do you use this skill?
After installing the toolkit in Claude Code: run /plugin-development:init my-awesome-plugin to scaffold a plugin; add components with /plugin-development:add-command|add-skill|add-agent|add-hook; run /plugin-development:validate after changes and /plugin-development:test-local to test in a dev marketplace. The plugin-authoring skill triggers automatically when context includes .claude-plugin/, plugin., hooks/, or similar paths, and stays read-only by default, asking before edits.