Back to directory

Agent Skills Platform

A tool that turns an existing workflow, described in a prompt plus supporting evidence such as spreadsheets or transcripts, into a tested, installable agent skill with evals and security checks, then governs its lifecycle through a marketplace with rollback and quarantine.

Visit website (opens in a new tab)francyjglisboa.github.io
Agent Skills Platform

Agent Skills Platform converts a real, already-followed workflow into an installable agent skill. Given a prompt describing the task plus the evidence behind it, such as spreadsheets, reports, emails, screenshots, transcripts, links or scripts, it builds a reusable skill package with instructions, any functional scripts needed, evaluations, security checks and a representative test run, that a person can inspect and an organization can review before others reuse it.

A skill can be created and used privately, with no marketplace involved, or published to a governed team marketplace hosted on GitHub or GitLab, where ownership and approval rules apply. After release, maintenance evidence from real use is captured separately from the skill's own runtime prompt, and recurring findings become draft patterns that only a separately validated change can turn into an update, rather than the skill silently modifying itself.

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.