Back to directory

Paperthin

Paperthin is a collection of small agent skills that turn established engineering habits, such as rewriting a drifted file cleanly or naming the one objection that could kill a plan, into steps a coding agent reaches for on its own.

Paperthin

Paperthin packages a set of low-level agentic design patterns as installable skills for coding agents such as Claude Code, Codex, Cursor and others. Each skill turns an established piece of engineering judgment, such as rewriting a drifted artifact cleanly instead of patching it further, checking whether a plan has one real objection that could kill it, or carving unsafe scope out of a task up front, into a reusable step the agent can trigger by name or invoke on its own.

Skills are organized by scope, by who invokes them, the model or the user, and by whether they only read without changing anything. Installing symlinks the skills into every configured agent so they stay current when the package updates.

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.