A comprehensive Claude Code plugin that assists developers in creating, validating, and distributing Claude Code plugins—from scaffold to release.
The plugin-development plugin provides a complete toolkit for plugin authorship using a hybrid architecture:
- Skill (plugin-authoring): Ambient, auto-discovered guidance with progressive disclosure
- 7 Slash Commands: Explicit actions for scaffolding, validation, and testing
- Reviewer Agent: Deep, multi-file audits and readiness checks
- Hooks: Automated validation and formatting guardrails
This plugin is part of the Claude Code Plugin Marketplace Template. To use it:
# From the template root directory
/plugin marketplace add .
# Install the plugin
/plugin install plugin-development@claude-code-plugin-templateIf you're using this template to create your own marketplace:
- Set up your marketplace (see the main README for instructions)
- Add your marketplace to Claude Code:
/plugin marketplace add your-org/your-marketplace-repo
- Install the plugin:
/plugin install plugin-development@your-marketplace-name
If you want to use this for developing new plugins:
# From the template root directory
cd /path/to/claude-code-plugin-template
# Add the marketplace
/plugin marketplace add .
# Install this plugin
/plugin install plugin-development@claude-code-plugin-templateThen use it to scaffold new plugins:
/plugin-development:init my-new-pluginBefore you begin: Make sure you've installed this plugin following the Installation instructions above.
Once installed, you can scaffold a new plugin:
/plugin-development:init my-awesome-pluginThis scaffolds:
.claude-plugin/plugin.json- Plugin manifestcommands/- Slash commands directoryagents/- Sub-agents directoryskills/- Skills directoryhooks/hooks.json- Hook configurationscripts/- Validation scriptsREADME.md- Documentation template
# Add a command
/plugin-development:add-command format-code "Format code according to project standards"
# Add a skill
/plugin-development:add-skill code-review "Use when reviewing code or PRs"
# Add an agent
/plugin-development:add-agent security-auditor "Analyzes code for security vulnerabilities"
# Add a hook
/plugin-development:add-hook PreToolUse "Write|Edit"/plugin-development:validateChecks:
- Plugin structure and organization
- Manifest validity (plugin.json)
- Component configuration
- Naming conventions
- Hook safety
After creating your plugin and making it available in a marketplace:
# From your plugin directory
/plugin-development:test-localThis creates a dev marketplace and provides installation instructions for iterative testing. See the Testing Your Plugin section below for complete workflow details.
| Command | Description |
|---|---|
/plugin-development:init [name] |
Scaffold a new plugin with standard structure |
/plugin-development:add-command [name] [desc] |
Add a new slash command |
/plugin-development:add-skill [name] [desc] |
Add a new Skill with SKILL.md |
/plugin-development:add-agent [name] [desc] |
Add a new sub-agent |
/plugin-development:add-hook [event] [matcher] |
Add a hook configuration |
/plugin-development:validate |
Validate plugin structure and configuration |
/plugin-development:test-local |
Create dev marketplace for local testing |
The plugin-authoring Skill provides ambient guidance throughout your development session.
Triggers: Automatically activates when working with:
.claude-plugin/plugin.jsonmarketplace.jsoncommands/,agents/,skills/, orhooks/directories
Capabilities:
- Read-only analysis and guidance
- Progressive disclosure (links to detailed docs)
- Proposes safe actions via commands
- Escalates to plugin-reviewer agent for deep audits
Access: Automatic (no explicit invocation needed)
Deep, comprehensive audits of plugins for release readiness.
Invoke: /agents plugin-reviewer
Provides:
- Structure audit
- Manifest validation
- Component analysis
- Hook safety checks
- Marketplace readiness assessment
- Prioritized issue list (Critical → Low)
- Specific fix recommendations
Runtime: 5-15 minutes (separate context window)
Automated validation and formatting hooks are included:
- Validates plugin structure before Write/Edit operations
- Exit code 2 blocks unsafe operations
- Provides actionable error messages
- Runs formatting/linting after Write/Edit (stub, extend as needed)
- Displays plugin load confirmation
Before starting plugin development:
-
Clone or download this template:
git clone https://github.com/your-org/claude-code-plugin-template.git cd claude-code-plugin-template -
Install this plugin (see Installation above)
-
Create a new plugin:
/plugin-development:init my-plugin cd my-plugin/
Once installed, create your plugin:
cd my-projects/
/plugin-development:init my-plugin
cd my-plugin//plugin-development:add-command hello "Greet the user"
# Edit commands/hello.md with your instructions
/plugin-development:add-skill my-expertise "Use when working with X"
# Edit skills/my-expertise/SKILL.md/plugin-development:validateFix any errors or warnings.
/plugin-development:test-local
/plugin marketplace add ../dev-marketplace
/plugin install my-plugin@dev-marketplaceTest your commands:
/my-plugin:helloAfter making changes:
/plugin-development:validate
/plugin uninstall my-plugin@dev-marketplace
/plugin install my-plugin@dev-marketplace
# Test again/agents plugin-reviewerFix any critical or high-priority issues identified.
Add to your team marketplace repository and push to GitHub.
The plugin includes comprehensive reference material:
schemas/plugin-manifest.md- plugin.json structureschemas/hooks-schema.md- Hooks configurationschemas/marketplace-schema.md- Marketplace structure
templates/plugin-manifest.json- Starter plugin.jsontemplates/marketplace-manifest.json- Marketplace templatetemplates/command-template.md- Command structuretemplates/skill-template.md- Skill structuretemplates/agent-template.md- Agent structure
examples/simple-plugin.md- Complete minimal plugin exampleexamples/testing-workflow.md- Step-by-step testing guide
best-practices/organization.md- Plugin structure guidelinesbest-practices/naming-conventions.md- Naming standards
Skill (Read-First):
- Ambient, automatic activation
- Read-only by default (Read, Grep, Glob)
- Proposes commands for execution
- Progressive disclosure keeps it concise
Commands (Explicit):
- User-triggered actions
- Deterministic operations
- Template-based generation
- Clear consent for modifications
Agent (Deep Analysis):
- Separate context window
- Multi-file analysis
- Comprehensive reporting
- Use for release readiness
Hooks (Automation):
- Validation before writes
- Formatting after writes
- Automated guardrails
- Safety first
- Skill is read-only by default
- All writes go through explicit commands
- Hooks block unsafe operations (exit code 2)
- Agent proposes changes, doesn't execute
For enhanced security, configure in .claude/settings.json:
{
"permissions": {
"deny": [
"Read(./.env)",
"Read(./.env.*)",
"Read(./secrets/**)"
],
"ask": [
"SlashCommand:/plugin-development:*"
]
}
}- Check
/plugin listto verify installation - Verify
plugin.jsondoesn't include component fields for standard paths - If using custom paths, ensure they're relative (e.g.,
["./custom/cmd.md"]) - Run
/plugin-development:validate
- Check
/helpfor command list - Ensure
commands/has.mdfiles with frontmatter - Verify
descriptionfield in frontmatter - Reinstall the plugin
- Check
SKILL.mdfrontmatternamematches directory (kebab-case) - Verify
descriptionincludes both what the Skill does AND when to use it - Include specific trigger keywords users would naturally mention
- Description must be max 1024 characters
- Skills are model-invoked (automatically activate based on context)
- Make scripts executable:
chmod +x scripts/*.sh - Use
${CLAUDE_PLUGIN_ROOT}in hook commands - Test script directly:
./scripts/validate-plugin.sh - Use
claude --debugto see hook execution
This plugin follows its own best practices:
- kebab-case naming throughout
- Progressive disclosure in skills
- Template-based generation
- Comprehensive validation
To extend:
- Add new commands in
commands/ - Enhance the skill with new reference docs
- Update the reviewer agent with new checks
- Add hooks for new automation
See skills/plugin-authoring/ for:
- Schemas (plugin manifest, hooks, marketplace)
- Templates (reusable starter files)
- Examples (complete working plugins)
- Best practices (organization, naming)
MIT License - See LICENSE file for details
1.0.0 - Initial release
- ✅ Plugin scaffolding (init command)
- ✅ Component generation (add-command, add-skill, add-agent, add-hook)
- ✅ Validation (validate command)
- ✅ Local testing (test-local command)
- ✅ Ambient guidance (plugin-authoring skill)
- ✅ Deep review (plugin-reviewer agent)
- ✅ Automated safety (validation hooks)
- ✅ Comprehensive documentation (schemas, templates, examples, best practices)
For issues or questions:
- Using the template: See the main README for template setup instructions
- Plugin issues: Check validation output:
/plugin-development:validate - Best practices: Review
skills/plugin-authoring/best-practices/ - Deep audit: Request deep audit:
/agents plugin-reviewer - Debug mode: Test with debug mode:
claude --debug