It captures every plan, request and decision you make.
New work attaches to the same plan instead of a new doc. Omino keeps track of what comes first, what has to wait and what can run in parallel, and your agents work from it.
Each task with a coding agent starts a new plan or spec, written for that one change. Weeks later you can't tell which came first, the same thing is described three ways, and an old doc the next agent reads may already be overridden.
One plan per project, shared by every agent and worktree. New work attaches to it: a stage, a lane, what it depends on. The agent reads it before acting and writes only when something changed, with your words and the source.
Without a command, the agent decides on its own each turn: decision or discussion, read only the notes involved, write only what changed, bring back ideas that are due, flag what needs a second look.
Commands are /omino-plan and so on in Claude Code, $omino-plan in Codex, and /omino:plan with the Claude Code plugin. Needs Node 18 or later.
Every worktree and clone shares one copy, so branches never conflict on it. The folder has its own git history, one commit per change. Nothing is added to your repo by default.
| Relation | Field | Blocks work? |
|---|
No recorded prerequisite does not mean two items can safely run in parallel, and omino never says it does.