About fluxly

A neutral, audited registry for agent skills.

Mission

Agent skills are scattered across dozens of repos and marketplaces, often with no security review and no consistent install path. fluxly aggregates them, runs a uniform audit, and publishes a single canonical install command per agent.

How we operate

How the index gets built

  1. 1 · Pull

    Every 24 hours we clone the upstream repos listed in scripts/sources.json, plus a curated list of GitHub repos that ship a SKILL.md.

  2. 2 · Audit

    gitleaks runs at entropy 3.8 and a separate scorer checks license, version, tool scope, install method and red flags. Each skill gets a 0–100 trust score.

  3. 3 · Translate

    Curated Chinese commentary takes priority; missing slots fall back to a paced Gemini call so every card carries a Chinese summary, fit/avoid hint and first-step.

  4. 4 · Publish

    The build emits one canonical install command per agent and rebuilds Cloudflare Pages on every push to main.

Where the content comes from

Anthropic's reference skills, Obra's superpowers, Matt Pocock's teaching repos, Baoyu (宝玉) Chinese-author skills, Garry Tan's gstack, Vercel labs, the UI/UX Pro Max community pack, plus dozens of public single-skill repos.

How to contribute

Open a PR against the upstream repo of any skill, or send us your own SKILL.md repo and we'll add it to the source list. Daily sync picks it up the next morning.

Contact

Bug reports and skill requests: GitHub issues. Security disclosures: security@fluxly.cc — we acknowledge within 48 hours.

Acknowledgements

Thanks to every author who ships SKILL.md publicly — the registry only works because you do. Special call-out to Anthropic, Obra, Matt Pocock, Baoyu, Garry Tan and the Vercel team for kicking off the ecosystem.