Back to directory

improve

An agent skill that audits a codebase with a capable model and writes self-contained implementation plans for cheaper models or other agents to execute.

improve

improve is an agent skill that audits any codebase and writes implementation plans for other agents to carry out. It never implements anything itself: a capable model handles the part where judgment compounds, understanding the codebase and deciding what is worth doing, and a cheaper model executes the resulting plan.

Running it maps the repository and returns a table of findings; replying with the ones worth pursuing produces one self-contained markdown plan file per finding, plus an index in a recommended order. A plan can be handed to any agent to implement, or the skill can dispatch a cheaper model to execute it in an isolated worktree and report back with a verdict, leaving the decision to merge to the user.

A later run can reconcile the backlog of plans, verifying what has landed and refreshing or unblocking what has not.

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.