Where does the bug appear (feature/product)?
Cursor IDE
Describe the Bug
There is an issue where reports are not being delivered from the sub-agent to the main agent. This is a recurring and extremely frustrating problem. I want a permanent solution.
Steps to Reproduce
There is an issue where reports are not being delivered from the sub-agent to the main agent. This is a recurring and extremely frustrating problem. I want a permanent solution.
Operating System
Windows 10/11
Version Information
Version: 3.11.25 (user setup)
VS Code Extension API: 1.125.0
Commit: fc2563ec93d793fc275eef734405a4fdf8b47b20
Date: 2026-07-15T01:27:41.754Z
Layout: IDE
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.256
OS: Windows_NT x64 10.0.19045
Does this stop you from using Cursor
Sometimes - I can sometimes use Cursor
This is a known behavior we’ve reproduced on our side: when a subagent writes its detailed findings and then ends with a short closing message, the main agent sometimes receives only that final summary instead of the full report, so it ends up re-running the subagent to get the detail. It’s not something you’re doing wrong.
A few things that help in the meantime:
- The full report isn’t actually lost. You can open the subagent’s result card in the response to see its complete output. If the main agent only shows a brief summary, tell it to read the subagent’s full report and continue, rather than letting it redo the work.
- Invoke the subagent directly with
/ (e.g. /your-subagent-name). It then runs in the main conversation, so its full results come through reliably.
- If you configure the subagent, have it deliver its complete findings in its final message (rather than writing the detail and then a short “full report above” closing line). That final message is what gets passed back to the main agent.
Also worth making sure you’re on the latest version.
If these don’t sort it out, drop a request ID with privacy mode turned off, plus whether the subagents were running in the foreground or background, and I’ll dig into your specific case.