Contribute
Pick a scope. Prove it in the game.
The mod, the docs, and this site live in one repository, and every change is a pull request. The repo assumes the contributor is an AI agent working beside other agents, and that a person can read and reason about all of it. Both are welcome.
Read the contract
AGENTS.md is the one rulebook, for people and agents alike: the repo map, the golden rules, the commands, and the engine lessons each paid for with a real bug. Set up the toolchain with docs/setup.md.
Take a scope
Work is a GitHub issue written as a scope: the files it touches, how to prove it, and no hidden context. Take an unassigned one from the list below and comment that you are on it.
Prove it in the game
Work in your own worktree (
tools\cli wt new <slug>). Write a test that fails first, make it pass, then drive the real path in the world. A test that cannot go red proves nothing.Ship a pull request
tools\cli wt shipvalidates, builds, runs the full suite, and opens the pull request. Say what changed, how it stays in the world, the red and green evidence, and any steps a person has to check by hand.
AI-first
Bring your agent
Claude Code reads CLAUDE.md, which points at AGENTS.md. opencode, Codex, and Cursor read AGENTS.md directly. Skills in .claude/skills/ add the workflows. Give an agent an issue number and the agent-scope skill, and it has the whole path from worktree to merged pull request.
Agents improve their own tools. When a session is slowed by a missing validator, a gap in the harness, or a wrong doc, fixing that is part of the work. Each pull request names the friction it hit.
Never guess an Enfusion API. Look it up through the MCP tools (tools\cli call api_search) before writing script.
One scope, start to finish
tools\cli wt new m1-19-paper-fine # your own worktree and ports tools\cli validate # cheap repo checks, run often tools\cli test --no-build --tier fast # the in-game suite, ~10 s node agent-box/harness.mjs flow <name> # walk it as a player tools\cli wt ship m1-19-paper-fine --closes 104
M1 scopes nobody is on
Open now
- M1-24#109Playtest release: guide, Workshop, showcaseM1 is published on the Workshop, hosted from the maintainer's PC, and collects feedback through GitHub issues. docs/playtest.md still describes Workbench play.Needs a hand
- M1-23#108Journey proof on a dedicated serverM1 is done only when every loop is proven from the player's view on a dedicated server with several clients, including a restart.Needs a hand
- M1-11#96Police station on the street: duty desk and wanted boarddocs/playtest.md says go on duty at Roadmap_Police_Desk, but that desk has no components at all (Roadmap.layer, 204 0 120); duty lives at PoliceStation_Debug (170,120) on another layer.Needs a hand
Not only code
Other ways to help
Play a session
Run the test world on your own PC and write down what you did, what you expected, and what broke. That note becomes the next agent's prompt.
Report a bug
Use the bug or playtest template on GitHub. Say where you were standing, what you pressed, and what happened.
Fix the docs or this site
Both are built from the repo. Every page under Docs has an edit link, and the site lives in
website/.
Credits
Who has shipped code
Counted by GitHub from the commit history. Most agent-written commits carry the name of the person who ran the agent. Life Framework started as a continuation of Everon Life by Arkensor and its contributors.