Autonomous goal-completion loop for AI agents. Set a goal once; the agent runs uninterrupted until done — on every tier-1 harness (Claude Code, Codex, OpenCode, Pi Mono).
agentplugins add sigilco/agentplugins-goalThe stop hook is the engine. When an agent session ends, hooks/goal-checker.js reads ~/.agentplugins/goal-state.json. If a goal is active and under the iteration limit, it returns { continueWith: "..." } to re-enter the agent automatically.
| Command | Description |
|---|---|
/goal <description> |
Set a goal and start the autonomous loop |
/goal-stop |
Mark the goal complete and stop the loop |
/goal-status |
Show goal, iteration count, and limit |
~/.agentplugins/goal-state.json tracks the active goal:
{
"active": true,
"goal": "refactor the auth module",
"iterations": 3,
"maxIterations": 20,
"startedAt": "2026-06-25T18:00:00.000Z"
}MIT