Back to directory

SkillCorpus

SkillCorpus is an open-source pipeline that turns scattered SKILL.md files from public repositories into a curated, retrieval-ready corpus of agent skills, with safety and license gating, quality evaluation, and task-specific skill matching built in.

SkillCorpus

SkillCorpus aggregates SKILL.md files from public source repositories, applies safety and license gates, evaluates their quality, and matches task-specific skills into an agent's context before it answers, so an agent has vetted procedural knowledge rather than an unvetted pile of files. A hosted SkillHub API and demo let a user try the result without cloning the repository, while the open-source pipeline itself can be pointed at a custom source registry, have its taxonomy, quality, deduplication, retrieval and export rules changed, or be self-hosted end to end with a user's own retrieval models and agent host.

It supports multi-source retrieval across local skills, EverMind SkillHub, ClawHub and skillhub.cn, with filtering, deduplication and a final selection step, and ships official plugins for hosts including OpenClaw, Hermes, Raven, WorkBuddy and DeepSeek Harness. A shared skills library lets several installed hosts read and register skills from one common local directory, with changes taking effect on the next turn.

Its core is Apache-2.0 licensed, with the matching and evaluation modules under MIT; each individual skill keeps its own upstream license.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.