Skip to main content
prompt GP ChatGPT

end_not_done — Claude Code prompt: what it does and how to install

The end_not_done prompt for Claude Code and ChatGPT handles tasks that cannot be completed fully or in part. The agent reports confirmed work, the exact blocker, and what is currently on screen.

Apache-2.0 ★ 9,488 updated 2026-11-16

Install into Claude Code · Manually: put the files in place All methods and platforms ↓
.claude/commands/end_not_done.md
Open on GitHub

What end_not_done does

The prompt sets the response format for a task that has no result or only a partial one. The agent states what it can confirm as done, what prevented further work, and what the screen shows now.

A user’s request to stop does not by itself call for this response. To install it in Claude Code, save the prompt as .claude/commands/end_not_done.md.

What it helps with

  • report a task that produced no result
  • separate completed work from unfinished work
  • state what blocked further progress
  • record what is currently on screen

Example prompts for the agent

  • The task could not be completed. Report what you confirmed, what blocked you, and what is on screen now.
  • Finish the part you can, then explain what remains and what prevented you from continuing.

How to install end_not_done

Manually: put the files in place

.claude/commands/end_not_done.md

Questions about end_not_done

How do I install end_not_done in Claude Code?

Save the prompt to .claude/commands/end_not_done.md.

What does end_not_done do?

It defines how the agent reports a task that failed or was only partly completed. The response covers confirmed work, the blocker, and the current screen state.

When should I use end_not_done?

Use it when the requested outcome cannot be reached or only part of the task can be done. A user asking the agent to stop is not, by itself, a reason to use this prompt.

Source text

end_not_done.md · Apache-2.0
End the task without the requested result, when it cannot be reached or only part of it can. In message, say what was confirmed done, what exactly stopped you and what the screen shows now. Do not use it just because the user asked you to stop.

© wangzan101 · Apache-2.0 · wangzan101/Mika

Similar prompts

All prompts for Agent setup: memory, context, orchestration →

From the same set132

The whole set →
phone_dialog This Claude Code prompt checks the text and visible buttons in the top phone dialog. Leave buttonText empty to inspect without clicking. GP ChatGPT ★ 9,488repo stars phone_back The phone_back Claude Code prompt presses Back on a phone’s virtual display. If the screen stays put, it uses the app’s visible back control instead. GP ChatGPT ★ 9,488repo stars web_upload The web_upload prompt for Claude Code sends workspace files to a file input on a page. It passes the control reference, snapshotId, and file paths. GP ChatGPT ★ 9,488repo stars knowledge_search The knowledge_search Claude Code prompt searches Mika’s knowledge about apps and websites. It narrows results by type and lets the agent read a matching entry. GP ChatGPT ★ 9,488repo stars web_navigate web_navigate is a Claude Code prompt that opens web pages in Mika’s built-in browser. The agent can then use tools to inspect and interact with the page. GP ChatGPT ★ 9,488repo stars handle_obstacle The handle_obstacle prompt for Claude Code and ChatGPT identifies obstacles on the current page or app. It gives a next-step hint for popups, sign-in screens, verification, access limits, or payment steps. GP ChatGPT ★ 9,488repo stars phone_gesture The phone_gesture Claude Code prompt describes touch gestures on a virtual display. It activates when an action needs a path from one or two fingers. GP ChatGPT ★ 9,488repo stars phone_launch_app The phone_launch_app Claude Code prompt opens an installed app on a phone’s virtual display using its exact package name. Screen-reading and control tools become available after the next call. GP ChatGPT ★ 9,488repo stars

Project sponsors

Companies that keep this analytics open