agent-toolkit

Minimalistic skills and rules for AI coding agents that assist your daily work in any software engineering project

Project-agnostic Works with any agent MIT
# install with one command
git clone https://github.com/eai-org/agent-toolkit.git && cd agent-toolkit && ./install.sh

On Claude Code it also adds a hook that pulls the repo and reruns your install once a day, so your skills and rules follow the repo. Turn it off with ./install.sh --no-auto-update.

Other ways to install: agentwheel · skills.sh · Claude Code plugin · README →

This toolkit is entirely open source and free to use. Give us a star on GitHub to support us.

The problem

Different projects, same repetitive tasks

  • Every project is a different world tech stacks, trackers, conventions, AI adoption
  • Yet the daily tasks repeat everywhere fetch a ticket, refine it, plan, review, reply
  • agent-toolkit lets AI assist exactly those generic skills that can be used in any project

"I kept rebuilding the same AI setup in every project, so I built one that works everywhere."

Principles

Core ideas behind every skill

Keep the context window sharp Atomic skills: one job each, nothing else loaded. why →
Offload to files, pick up fresh Every phase ends in a doc a new session can pick up, and a fresh session can check it. how →
Human in the loop The agent recommends, you decide, and it explains things so you get them on the first read. Nothing runs behind your back.
Never guess It reads the existing code first and asks you when in doubt.
Learn from mistakes Every correction, and every review finding you accept, becomes a durable lesson. more →
Versatile by design Any project, any stack, team or solo.
Philosophy

A toolkit, not a framework

Great frameworks exist that auto-activate around everything you do. This one deliberately takes the other road:

auto-activating frameworks
  • a whole methodology
  • the agent drives
  • fuller context, less to remember
agent-toolkit
  • use only the parts you need
  • you drive, the agent assists
  • lean context, predictable behavior
Opinionated rules

Optional and not installed by default: get them all with ./install-opinionated-rules.sh or pick only the ones you want. Examples:

  • git-read-only-by-default no commits, pushes or resets unless you asked for them
  • no-ai-attribution your work stays yours: no AI co-author, no "generated with" footer
  • no-nonsense-comments only comments a future reader with zero context still needs

Not a native English speaker? One more rule makes your agent talk to you in your own language, while everything it writes into the project stays in English. Get the use-my-mothertongue rule →

Check the full list of available rules →

Share your feedback

Got an issue or an idea? Please report it on GitHub.

Open an issue →