Plugins Claude Code

Distribution format bundling skills, subagents, hooks, MCP servers, and LSP servers into an installable package. Plugins allow sharing or selling a complete Claude Code workflow, one install command deploys the entire configuration. Distributed via marketplaces (official Anthropic or third-party). Example: Ralph Wiggum Loop is a plugin. When to choose what: a standalone .skill.md file is enough for a personal workflow or a single project (fast, no versioning); a plugin becomes necessary as soon as you share with a team, an organization or through a marketplace, it adds the bundle layer (skills + agents + hooks + MCP) with versioning. What the word plugin does not tell you is what sits inside, and that changes everything in use. Three plugins published by Anthropic, often shown in the same list, make the point. frontend-design is a skill: guidance the agent reads when it works on an interface. code-review is a command: it runs four agents in parallel on a pull request and keeps only the issues scored 80 out of 100 or higher. security-guidance is a set of hooks: they warn while files are edited, reread the diff at the end of each turn and at commit time, then hand their findings back to Claude to fix. That last one is regularly sold as a security scan before deployment, which it is not. Read the list of components inside a plugin before trusting what it promises.

Strengths

Limitations

Best for

Official site

View on Coeurdar