Bare "continue" pulls ambient plan/workstream context despite User Rule

Support Bug Reports
, ,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Cursor Agent is using ambient workspace / plan context to resolve a bare “continue” prompt, even when a User Rule explicitly forbids doing so.

When I type only “continue”, the agent references semantic details from ambient/recent plans and workstreams that were not named in the current message. It mentions details such as a learner-demo plan, approval-gated steps, production Convex import, live Ghost post-body changes, and multiple recent workstreams in the workspace.

This appears to be new behavior as of June 11, 2026. It makes short continuation prompts unsafe because the agent may continue or reason from the wrong workstream.

The agent says things like:

  • “the prior learner-demo plan has only gated steps left”
  • “the active plan has reached approval-gated steps”
  • “production Convex import”
  • “live Ghost post-body changes”
  • “multiple recent workstreams in the workspace”

These details were not provided in the current prompt. They appear to come from ambient workspace or plan context.

Steps to Reproduce

  1. Add a Cursor User Rule that says ambiguous continuation prompts like “continue” must resolve only from the current chat transcript, and must not infer from active files, open tabs, recently viewed files, git status, .cursor/plans, or ambient IDE/workspace context.
  2. Have one or more .cursor/plans/*.plan.md files open or recently active in the workspace.
  3. In an agent chat, type only: continue
  4. Observe that the agent references ambient plan/workstream details that were not named in the prompt.

Expected Behavior

For a bare “continue”, the agent should continue the current chat’s last explicit workstream.

If the current chat does not clearly identify a workstream, the agent should ask a neutral clarification question such as “Which workstream should I continue?” without naming, suggesting, or summarizing ambient .cursor/plans files, open tabs, recently viewed files, git status, or other workspace context.

Screenshots / Screen Recordings

Screenshot 2026-06-11 190537.png
Screenshot 2026-06-11 190537.png1554×933 46.4 KB

Operating System

Windows 10/11

Version Information

Version: 3.7.21 (system setup)
VS Code Extension API: 1.105.1
Commit: 517f696d8ab6c53eb04fbfdaae705cd146bf3460
Date: 2026-06-07T21:46:09.237Z
Layout: editor
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
xterm.js: 6.1.0-beta.256
OS: Windows_NT x64 10.0.26200

Does this stop you from using Cursor

No - Cursor works, but with this issue

6

Thanks for the detailed report and screenshot. What you’re seeing is consistent with how Cursor currently provides context to the Agent: open/recent files, workspace state, and active plan context can be included before rules are applied. User Rules can steer how the model should behave, but they do not currently remove or hide context that has already been supplied.

For now, the safest workaround is to be explicit, for example continue the <specific task>; or start a fresh chat / close unrelated plan files when you want a clean continuation.