// the find
anthropics/claude-plugins-official
Official, Anthropic-managed directory of high quality Claude Code Plugins.
The official Anthropic-managed registry of Claude Code plugins — a monorepo of slash commands, agent definitions, skills, and MCP server wrappers that extend what Claude Code can do in your editor. It's the app store equivalent for Claude Code power users and the canonical reference for anyone building their own plugin. If you're already using Claude Code daily, this is the first place to look before rolling your own workflow tooling.
The plugin manifest schema is well-defined and the example-plugin directory gives you a working skeleton with every optional section present — faster to read than the docs. The external PR scope guard plus the SHA-pinning workflow for third-party plugins is the right call; it means a compromised upstream repo can't silently change what gets installed. The internal plugins cover the non-obvious use cases well: hookify (conditional pre/post-tool hooks), code-modernization (multi-agent legacy rewrite pipeline), and the LSP wrappers (clangd, gopls, jdtls, kotlin) that unlock proper diagnostics without any user config. The skill-bundle mechanism for repos that ship SKILL.md files without a full manifest is a pragmatic escape hatch.
The security disclaimer in the README — 'Anthropic does not control what MCP servers, files, or other software are included in plugins' — is honest but undersells the actual risk; external plugins run arbitrary code and the scan-plugins workflow isn't described anywhere in the repo, so you can't evaluate how much that gate actually catches. The external plugin list is thin (15-ish entries), making this feel more like a first-party showcase than a real marketplace. There's no versioning story for plugins beyond SHA pinning — if a skill changes behavior, existing installs get the update silently on next sync with no changelog. The Python-heavy hookify plugin ships its own rule engine and config loader rather than using Claude's existing hooks JSON format, which means two parallel extension points that overlap in purpose.