Frustrating prompt updates thisweek?

Ideas Feedback
, , ,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Starting this week the models are again not listening to plan and not get ahead of themselves. In debug mode especially it flies unattending with debug/fix cycles regardless of my prompts saying ask before proceeding. Last week/week before opus-4.6 was an amazing model. This week its been infuriating. not following instructions (mine), getting ahead of itself, and being too eager to make changes. please review the system prompt changes and revert or refine.

Steps to Reproduce

Work with anything and watch it rage ahead even when you say do not do anything without approval.

Expected Behavior

Not be so damn eager to make changes without first talking about it. Really ought to have a cursor IDE mode prompt and a standalone agent prompt where the cursor prompts understand there’s a human involved.

Operating System

MacOS

Version Information

Version: 2.5.25
VSCode Version: 1.105.1
Commit: 7150844152b426ed50d2b68dd6b33b5c5beb73c0
Date: 2026-02-24T07:17:49.417Z
Build Type: Stable
Release Track: Early Access
Electron: 39.4.0
Chromium: 142.0.7444.265
Node.js: 22.22.0
V8: 14.2.231.22-electron.0
OS: Darwin arm64 24.6.0

For AI issues: which model did you use?

Opus-4.6

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

3

Hey, thanks for the report. It’s really frustrating when an agent behaves incorrectly.

Here are a few things to check or try:

  1. Auto-run setting: make sure it’s turned off: Cursor Settings > Agents > Auto-run. When it’s off, the agent should ask for confirmation before running terminal commands.

  2. Cursor Rules: if you’re using .cursor/rules to tell the agent to “ask before continuing”, it works as a strong hint, but it’s not guaranteed. Models sometimes ignore rules like this, especially during debugging loops. Try making the rule as explicit as possible, for example: “NEVER make changes without my explicit approval. After suggesting changes, STOP and WAIT for my reply.”

  3. Debug mode in particular often behaves more aggressively because the agent enters a fix apply test loop. This is a known issue.

Can you share:

  • Your current auto-run setting (on or off)
  • Whether you have any .cursor/rules files
  • The Request ID from the session where this happened (Chat context menu > Copy Request ID)

That’ll help us figure out whether a specific change affected Opus-4.6 behavior. The team is aware of the general issue of agents being too proactive, and your report helps with prioritization.

Let me know how it goes.