f/prompts.chat's Claude Code Plugin for Prompts and Skills

What does the f/prompts.chat plugin do?

f/prompts.chat is the Claude Code plugin for prompts.chat, a self-hosted library of AI prompts and Agent Skills that has picked up 167,890 GitHub stars. Flowy files it under content, and the plugin's own summary is direct about the job: it searches and retrieves prompts or skill definitions from that library, and it runs whenever a session invokes the prompts or skills command in a bash or docker environment. Seven components ship with it in total: two agents, two commands, one MCP server, and two skills that activate on their own when a request matches.

Why does it have two of almost everything?

The plugin covers two separate kinds of content on prompts.chat, prompts and Agent Skills, and gives each one the same three piece kit: a command you type, an agent that does the fuller work, and a skill that activates itself.

On the prompts side, the prompts command searches and discovers AI prompts from prompts.chat. The prompt-manager agent goes further: it searches, saves, improves, and organizes a prompt library rather than only finding entries in it. The prompt-lookup skill activates on its own whenever a request asks about AI prompts, needs a prompt template, asks to search for prompts, or mentions prompts.chat by name, and it exists to help with discovering, retrieving, and improving a prompt once found.

On the Agent Skills side, the skills command searches and discovers Agent Skills from prompts.chat. The skill-manager agent searches, creates, and manages multi-file skills for Claude Code, so it can produce a new skill and not only locate an existing one. The skill-lookup skill activates when a request asks to find skills, browse a skill catalog, install a skill for Claude, or otherwise extend Claude's abilities with a reusable component. Its own description names installing a skill as one of its jobs, not just finding one, which makes it the piece that turns a search into something a session can actually use.

What ties the two tracks together?

One MCP server sits under both tracks, and the fact pack names it prompts.chat, the same name as the library itself. Its own one-line description is blank in the fact pack, and the plugin's public repository does not break its behavior down component by component either. What the repository does say is that this MCP server is how outside tools reach the library programmatically, rather than a person browsing the website by hand. skill-lookup's own description backs that up, saying plainly that it works "using MCP tools". Treat that as the shape of the MCP server's job, not a full account of it.

When would this plugin actually help?

Reach for it in either of two moments: you want a prompt to start from instead of writing one cold, or you want an existing Agent Skill instead of authoring one from scratch. Both agents go past simple lookup once you are there. prompt-manager can save and improve a prompt, and skill-manager can create a brand new multi-file skill. If neither moment applies, the two commands and two auto-invoked skills stay out of the way.

How do I install it?

Flowy indexes f/prompts.chat from its public GitHub repository. It doesn't run the prompts.chat service, and it doesn't host, own, or bundle the plugin. It points to the source and keeps the component list current. Install from the listing page linked at the top of this guide, and its two agents, two commands, one MCP server, and two skills become available in that session.