> ## Content Index > Fetch the complete content index at: https://www.notatechguy.com/llms.txt > Use this file to discover other available public pages before exploring further. # Claude Skills directory hits 75,000 GitHub stars - URL: https://www.notatechguy.com/claude-skills-directory-hits-75-000-github-stars/ - Published: 2026-09-18T15:42:22.000Z - Updated: 2026-09-18T15:42:22.000Z - Description: A community-maintained directory of Claude Skills has crossed 75,000 GitHub stars, but its production-ready claim is a self-report, not an audit. - Author: Marcello Babbili - Tags: Technology & AI, Anthropic, OpenAI ComposioHQ's awesome-claude-skills repository crossed 75,000 stars on GitHub today, trending on the platform's daily list with 70 new stars [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). The repo claims to collect more than 1,000 "production-ready" Claude Skills and plugins, a figure that comes from the maintainers' own README, not an independent audit [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). **My read:** This is the largest Claude Skills directory by star count, and the growth is real: the repo was created in October 2025 [P²](https://github.com/ComposioHQ/awesome-claude-skills?tab=readme-ov-file&ref=notatechguy.com) and has outpaced every comparable list. But the "1000+ production-ready" claim is a self-report from ComposioHQ, a company that also sells an MCP Gateway product, so the directory doubles as a distribution channel for their own integration platform. I'd treat the skill count as an upper bound until someone verifies the list independently. The 1,367 open issues on the repo [P²](https://github.com/ComposioHQ/awesome-claude-skills?tab=readme-ov-file&ref=notatechguy.com) suggest not everything in the list is working as advertised. ### How a skill is structured and loaded A skill lives in a folder built around a SKILL.md file. That file pairs YAML frontmatter — holding a name and description, with Markdown instructions, and may also include scripts and reference files [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). Skills are lazy-loaded: at session start the agent only reads each skill's name and description, costing roughly 100 tokens (the unit AI models use to measure text) per skill. The full SKILL.md content, typically under 5,000 tokens, is pulled in only once the agent determines that skill is relevant to the current task [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). The README separates three concepts. MCP (Model Context Protocol) governs the connection layer — how an agent authenticates to, transports data to, and discovers tools on external systems. Tools are the discrete functions an agent invokes. Skills sit above both, encoding the workflow: the ordered steps that tell the agent which tools to call and in what sequence to finish a job [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). ### The maintainer and the commercial tie-in The repository is maintained by ComposioHQ, not Anthropic [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). The README states that Anthropic first published the Claude Skills format in October 2025 and opened it as a standard in December 2025, though these dates are the maintainers' own account and have not been independently confirmed [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). The same README lists support across Claude Code, Claude.ai, the Claude API, OpenAI Codex, Cursor, Gemini CLI, Antigravity, and Windsurf [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). ComposioHQ also uses the repository to promote its Composio MCP Gateway, advertising one MCP endpoint that covers 1,000+ integrations with built-in authentication and audit logging [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). A connect-apps plugin shipped inside the repo routes Claude through Composio's infrastructure to carry out real actions across 1,000+ apps [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com). The directory and the commercial product feed each other. The comparable directories are smaller. Hugging Face's own skills repo, created in November 2025, sits at 10,955 stars [P³](https://github.com/huggingface/skills/?ref=notatechguy.com). Travisvn's awesome-claude-skills list, created a day before ComposioHQ's, has 14,633 stars [P⁴](https://github.com/travisvn/awesome-claude-Skills?ref=notatechguy.com). ComposioHQ's repo leads both by a wide margin, with 75,167 stars and 8,493 forks as of the latest snapshot [S¹](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com)[P²](https://github.com/ComposioHQ/awesome-claude-skills?tab=readme-ov-file&ref=notatechguy.com). ![GitHub stars: Claude Skills directories](https://storage.ghost.io/c/6e/89/6e896869-22ef-4281-a213-b4c462c17cff/content/images/2026/09/chart_f7906b1edd0fec630a2e.png) For a developer evaluating this list, the practical move is to spot-check individual skills before adopting them. The "production-ready" label covers everything in the directory, but nobody outside the maintainers has verified that all entries meet a consistent quality bar. A team building on Claude Code could start with the skills most relevant to their stack, read the SKILL.md files directly, and test them against a known task before committing. For a DevOps lead who maintains a Claude Code deployment, the immediate change is a new vetting step: before any skill from this directory lands in the team's config, someone on the desk needs to open the SKILL.md, confirm the tool calls it references actually exist in the team's MCP setup, and run it against a sandbox task — because the directory's quality bar is self-assigned and the 1,367 open issues indicate real breakage. --- *Sources: [S1 — ComposioHQ/awesome-claude-skills: A curated list of awesome Claude Ski](https://github.com/ComposioHQ/awesome-claude-skills?ref=notatechguy.com) · [P2 — ComposioHQ/awesome-claude-skills](https://github.com/ComposioHQ/awesome-claude-skills?tab=readme-ov-file&ref=notatechguy.com) · [P3 — huggingface/skills](https://github.com/huggingface/skills/?ref=notatechguy.com) · [P4 — travisvn/awesome-claude-skills](https://github.com/travisvn/awesome-claude-Skills?ref=notatechguy.com) · [P5 — VILA-Lab/Dive-into-Claude-Code](https://github.com/VILA-Lab/Dive-into-Claude-Code?ref=notatechguy.com)* --- *Written from 5 sourced items, 4 of them primary.* ## More from Not A Tech Guy - [NYU 7.4B model matches GPT-3 13B with 20× less compute](https://www.notatechguy.com/nyu-7-4b-model-matches-gpt-3-13b-with-20-less-compute/) - [OpenAI tells firms to measure AI ROI with its own tools](https://www.notatechguy.com/openai-tells-firms-to-measure-ai-roi-with-its-own-tools/) - [OpenAI previews Sponsored Agents in ChatGPT ads push](https://www.notatechguy.com/openai-previews-sponsored-agents-in-chatgpt-ads-push/)