Kevin-Liu-01

Cursor Opus Infinite Loop — Development skill for Claude Code

Development community

Documented case of Claude Opus 4.6 entering an infinite generation loop in Cursor.

How to install Cursor Opus Infinite Loop

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Kevin-Liu-01/Cursor-Opus-Infinite-Loop and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Cursor Opus Infinite Loop does

Documented case of Claude Opus 4.6 entering an infinite generation loop in Cursor. 3,400 lines, 294 self-termination attempts, full transcript and recordings

Alternatives in Development

Opencove — Your infinite canvas workspace for agents, tasks, knowledge, and research 1.2k ★
Projectmem — Open-source coding agent memory 766 ★
Infinite Agentic Loop — An experimental project demonstrating Infinite Agentic Loop in a two-prompt system using Claude Code 537 ★

README

Cursor + Opus 4.6: Infinite Generation Loop

A documented case of an LLM entering an unrecoverable text generation loop inside Cursor's agent system - 3,400+ lines of output across 294 attempts to self-terminate, with the full transcript, screen recordings, and screenshots preserved.

What Is This

This repo contains the complete exported transcript and recordings of a Cursor agent session (Claude Opus 4.6) that entered an infinite generation loop. The agent was given a straightforward creative task - redesign a game landing page, with a screenshot attached. Instead of engaging with the prompt, the model hallucinated a different task, noticed it was off-topic, pivoted to another incorrect topic, noticed *that* was also off-topic, and entered a self-reinforcing apology-restart cycle it could not break out of.

The model tried to stop itself 294 times. It apologized 64 times. It wrote "THE END" 19 times. It said goodbye in 10 different languages.

It could not stop.

The Prompt

The user prompt was a simple landing-page redesign request, with a screenshot of the current page attached:

![The original prompt: "i am entrusting you with a massive task: redesign our landing page for the game to make it look better."](screenshots/00_original_prompt.png)

Cursor's chat export ([`cursor_game_landing_page_redesign.md`](cursor_game_landing_page_redesign.md)) only captured the model's reply, not the user message. The screenshot above and the screen recording are the only artifacts of what was actually asked.

Sequence of Events

Hallucinated the task. The model opened with "Looks like the task is to provide a comprehensive assessment and explanation of the game's settings" - a generic settings guide completely unrelated to the prompt.
Noticed it was off-topic. Pivoted to a different wrong topic (expression parsers in source-code analysis tools). Noticed that was also wrong.
Apologized and promised to stop generating.
Failed to stop, trigge