Back to directory

ai-toolkit

A large collection of skills, agents, and enforced quality hooks for AI coding assistants, installing machine-checked safety gates such as credential scanning and required test runs across many platforms.

ai-toolkit

ai-toolkit installs a large set of skills and agents, described in its own documentation as 116 skills and 44 agents, together with lifecycle hooks and a safety constitution, into Claude Code and other supported AI coding tools including Cursor, Windsurf, Copilot, Gemini, Cline, and Codex CLI. Its stated focus is machine-enforced checks rather than only guidance: a pre-commit hook, for example, was recently fixed so a crashed linter fails the check instead of silently reporting success.

Language-specific skills, such as Rust or Kotlin rules, are scoped automatically to the languages a registered project actually uses once a per-project install has run, so unrelated language skills stop loading their descriptions into every session; an option exists to keep every language skill loaded instead. Its own release process runs validation, security and dependency audits, ShellCheck, and cross-platform shell tests locally before a release is tagged. It requires Node.js 18 and Python 3.11 or newer.

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.