pi-codex-goal plugin
A Pi extension that adds a Codex-style persistent goal so the agent continues across turns until the goal is done.
- 207 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Auto-continuation starts new agent turns and spends tokens without separate confirmation of each.
Install
Manual install
pi install npm:pi-codex-goalThe first command of Install and start installs the extension from npm into Pi. For an already open Pi add `/reload`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Auto-continuation starts new agent turns and spends tokens without separate confirmation of each.
This is third-party code. Review the repository files before installing.
What it does
pi-codex-goal adds a persistent goal to Pi and prompts the agent to keep working on it in later turns. The `/create-goal` command expands your task into a statement with completion checks, and `/goal` shows the objective, status, token budget, usage and active time. A new goal replaces the old one. The `/goal pause`, `/goal resume`, `/goal resume cancel`, `/goal copy` and `/goal clear` commands control progress, and Esc stops the current turn and pauses the active goal. Follow-ups generated by the goal arrive as hidden service messages rather than editable input. Work stops on completion, pause, a blocker or the token limit. The extension supports official Pi releases and the fitchmultz fork.
Who it is for. Pi users who want long tasks without constantly telling the agent to continue.
Good fit when
- A task spans many turns and the agent should come back to it by itself.
- A goal statement with completion checks is needed instead of a vague request.
- Work must be paused and later resumed after changes to the plan.
- Token spend and time spent on the goal should be visible.
Not a fit when
- You do not work in Pi, and the extension does not fit Claude Code or Codex directly.
- The task is short and closes in one turn.
- The token budget is tight and auto-continuation is unwanted.
Example request
/create-goal Add CSV export to the reports page. Test the export. Document its use.Limitations
The extension works only inside Pi, and Node.js and Pi version requirements are described in `docs/reference.md`. Auto-continuation spends tokens, so watch the goal budget and pause if the direction changes. A new goal replaces the current one, so include every requirement you want to keep in `/create-goal`. The extension does not improve the model itself: it only reminds the agent to continue. The license is MIT. The last commit is dated November 2026.
How to disable. Remove the extension through the Pi package manager and run `/reload`. Clear the current goal first with `/goal clear`.
FAQ
How do I install the extension?
Run `pi install npm:pi-codex-goal`, then start `pi`. If Pi is already open, call `/reload`.
How does `/create-goal` differ from `/goal`?
`/create-goal` expands a task into a goal with completion checks. `/goal <objective>` sets the exact wording without expansion.
What does the Esc key do?
It stops the current turn and pauses the active goal. Continue with `/goal resume`.
When does the goal stop?
On completion, pause, a blocker or reaching the token limit. `/goal` shows spend details.
Can I cancel the retry after a limit?
Yes, `/goal resume cancel` cancels a scheduled retry after a provider limit.
Discussion
No questions or reviews yet.
Related
PluginsA skill that builds a Codex /goal command from a five-section template so a long task does not drift and passes the audit
A meta-skill for Codex: a vague task becomes a ready "/goal" command with verification, boundaries and pause conditions.
CodexPotter loops Codex with a clean context each round until the code matches the goal, as a replacement for the built-in /goal.
A skill that turns a vague request into a Goal Prompt with success criteria and stop conditions, plus a Loop Prompt if needed.
A minimal terminal coding agent extended with skills, prompt templates and TypeScript packages
Personal Pi agent setup with four commands for plan, todos, execution and review, passing state through Markdown session files.
README in short
The README covers installation, a table of goal control commands and the behavior of service messages. Installation details and the reference live in `docs/reference.md`.