claude-code-plugin-dev

Build Claude Code plugins by structuring your project with manifests following the 2025 schema, implementing slash commands and agent skills with tool permissions, then testing locally before publishing. The plugin architecture guides you through directory layout, manifest configuration, and activation triggers needed for deployment.

Slash commands and agent skills extend Claude Code's interface with custom actions; MCP servers expose tools and resources through the Model Context Protocol; hooks trigger on specific events. All three plugin types share manifest schemas and distribution workflows but serve different extension points in the Claude Code ecosystem.

Test plugins locally using development workflows outlined in the schema, validate manifests against the 2025 specification, check tool permissions and version tracking, then run activation triggers to confirm behavior. Security best practices and testing procedures ensure reliability before distributing to the plugin registry.

Yes, you can build and deploy Claude Code plugins internally by implementing the manifest structure and schema, setting up your directory layout, and configuring activation triggers for your environment. The end-to-end workflow supports both internal distribution and public publishing pathways.

Plugin manifests declare tool permissions explicitly in the 2025 schema and include version tracking to manage updates and compatibility. These configuration elements control what actions your slash commands, agent skills, MCP servers, and hooks can perform within Claude Code and how they evolve over time.

Structure plugins with manifests following the 2025 schema, organize source files by component type (slash commands, agent skills, hooks, MCP servers), implement testing procedures, and prepare publishing documentation. This architecture streamlines development workflows and ensures your plugin meets security and technical requirements for distribution.