Back to directory

rnskill

A collection of more than sixty AI Agent Skills for Codex and Claude Code, spanning social-media content creation, such as scripting and captioning, and general coding skills.

rnskill

rnskill is a maintained collection of more than sixty top-level Agent Skills, plus nested skill sets, for coding agents that support the SKILL.md format such as Codex and Claude Code. It groups skills into two areas: content creation for video, graphics and articles, and general-purpose coding skills that are continually expanded.

Two of its content skills work together in sequence: one reads novel chapters and picks out a single event that stands on its own, sorting original facts from suggested audiovisual treatment and new adaptation, and outputs an event card and shot list; the other then builds continuous reference keyframes from a confirmed style image for a character and its larger stylized form appearing together. A separate skill checks whether a report, proposal, spreadsheet or slide deck's cited sources actually support its conclusions before it is delivered, flagging items for human review; it needs Python and an API key for its live checks, or can preview requests and review saved responses offline. Skills install through a skills command-line tool or a Claude Code plugin marketplace.

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.