Where does the bug appear (feature/product)?
Cursor IDE
Describe the Bug
The in-agent AskQuestion tool (the structured multiple-choice question card the assistant shows in chat) does not reliably deliver the user’s submitted answers back to the agent. Two distinct failures happened in a single session (agent in Multitask Mode, Remote-SSH workspace):
-
An AskQuestion card was answered and submitted by the user, but the agent received the tool result “AskQuestion was interrupted by the user before it completed” — i.e. NO answers were delivered. The user said “I answered your questions” while the agent only saw an interruption, and had to re-ask the whole card.
-
On a later AskQuestion that included a free-text “Other” answer, the text delivered to the agent DIFFERED from what the user actually typed: the user entered “treat it as cash” but the agent received “treat it as bank payment”. The agent then implemented the wrong behavior based on the incorrect relayed answer, which had to be caught and undone.
Net effect: the agent acts on missing or incorrect answers — wasted work, rework, and user frustration.
Steps to Reproduce
Intermittent. In an agent chat (Multitask Mode, over a Remote-SSH workspace):
- The agent uses the AskQuestion tool to present a multiple-choice question card, including at least one question with a free-text “Other” option.
- The user selects answers (and types a custom “Other” answer), then submits the card.
Observed failures:
- (a) Sometimes the agent receives “AskQuestion was interrupted by the user before it completed” even though the user actually submitted answers (no answers delivered).
- (b) Sometimes a free-text “Other” answer arrives at the agent as different text than the user typed (observed: user typed “treat it as cash”, agent received “treat it as bank payment”).
Expected Behavior
The exact option(s) selected and any free-text “Other” content the user submits in an AskQuestion card should be delivered verbatim to the agent, and a completed submission should never surface to the agent as “interrupted”. Please verify the AskQuestion answer round-trip (UI submit → agent tool result), especially (a) the “interrupted” state being reported despite a real submission, and (b) free-text answer fidelity.
Operating System
MacOS
Version Information
Cursor desktop (macOS), Remote-SSH to a Linux EC2 workspace, ~2026-06-16. Exact build: Version: 3.7.42
VS Code Extension API: 1.105.1
Commit: 5702c9cfca656d8710fad58402fe37f14345e3a0
Date: 2026-06-15T19:39:42.738Z
Layout: glass
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: Darwin arm64 25.5.0
Additional Information
Workspace is Remote-SSH from a macOS Cursor desktop client to a Linux (EC2) host. Agent running in Multitask Mode with background subagents. The failure has happened more than once in the same session, both with the false “interrupted” result and with a garbled free-text “Other” answer.
Does this stop you from using Cursor
No - Cursor works, but with this issue