MCP App Store
by KeenEthics
Productivity
LLM SkillHub icon

LLM SkillHub

by LLM SkillHub

Overview

LLM SkillHub is a catalogue of 600+ skills for AI agents: SKILL.md packages of instructions, sometimes with scripts, that teach an agent a specific job, plus evals that measure how well a skill does it. Skills work in Claude Code, Cursor, Codex, Gemini CLI and OpenCode, and load only when a task needs them. Every public version is scanned for leaked credentials and prompt injection, then read by a person before it is listed. Use it when you want a ready-made way to do a task instead of writing the instructions yourself, for example: "find a skill for reviewing pull requests", "is there a skill to triage CI failures?", "a skill that writes cold outreach emails", or "check the internal links in my markdown docs". With this connector, Claude can: - Search the catalogue by describing a task in plain words. Search matches meaning, not just keywords. - Look up a skill by name, with its licence, categories and every published version. - Read any file in a published version, starting with SKILL.md, so you can see exactly what a skill will do before you use it. - Browse the catalogue by category. - Get the exact command to install a skill with the free llmsh CLI. Every tool is read-only. The connector never installs, publishes or changes anything: the install tool returns a command for you to review and run yourself. No account is needed to read the public catalogue, and searches are not stored.

Tools

get_skill

Claude
Full detail for one named skill: description, categories, licence, and every published version with its digest and size. Use whenever you already know the name — from the user, from a link, or from a search_skills result — rather than searching for a name you have. Searching for a name you already know costs a call and can return something else.

install_command

Claude
The exact command to install a skill. Show it to the user and let them run it — this does not install anything itself. Pass agent with the agent you are running in, so the skill lands where that agent loads skills.

list_categories

Claude
The catalogue's category vocabulary, with how many skills are in each. Use to browse when a search term is not obvious.

read_skill_file

Claude
Read one file out of a published version. Defaults to SKILL.md, the skill's own instructions, so a name alone answers "what does this actually do" — you do not need get_skill first. Pass path for any script or reference it ships.

search_skills

Claude
Find skills by describing the task, in a sentence, in whatever words the user used — the catalogue matches meaning as well as words, so "make an animated picture for chat" finds a Slack GIF skill that shares none of those words with it. Use when the user asks for a capability you do not have, or describes something a published skill might already do. If you already know the skill's name, call get_skill instead: this searches, that fetches. Read the first line of the result: it says whether these matched what you asked or are only related to it. Each result then has a match line. high: it was named, or contains every word asked and matches by meaning -- a fair answer to recommend. medium: one kind of evidence -- read it with get_skill before recommending. low: related by meaning or by a few words only -- not an answer; say so, or keep looking.

App Stats

5

Tools

Claude

Platforms

Works with

Claude

Data refreshed daily