[bug] GitHub Copilot: Raptor mini (Preview) fails with “Sorry, no response was returned” when `execute` tools are enabled

[singularitti]

• Copilot Chat Extension Version: latest builtin
• VS Code Version: 1.119.0 (commit 8b640ee), date 2026-05-05T11:23:50-
• OS Version: macOS 26.4.1 (25E253)
• Feature (e.g. agent/edit/ask mode): Agent mode
• Selected model (e.g. GPT 4.1, Claude 3.7 Sonnet): Raptor mini (Preview)

Steps to Reproduce:

1. Open a completely empty VS Code window.
2. No project/workspace is required.
3. Open Copilot Chat in Agent mode.
4. Select the Raptor mini (Preview) model.
5. Enable the following execute tools:
   • createAndRunTask
   • getTerminalOutput
   • killTerminal
6. Send a very simple prompt such as: 1+1=?
7. The model immediately fails with:

   I'm sorry, but I cannot assist with that request.
   Sorry, no response was returned.

8. Click the “Try again” button.
9. Observe the same failure.
10. Restart VS Code.
11. Observe the same failure.
12. Reinstall the extension and restart VS Code.
13. Observe the same failure again.
14. Disable any one of the three tools above.
15. The model immediately starts working normally again.

Expected Behavior

Raptor mini (Preview) should function normally when the execute toolset is enabled, including when:

• createAndRunTask
• getTerminalOutput
• killTerminal

are all enabled simultaneously.

Additional Context:

• Similar issues have already been reported multiple times on GitHub involving tool-related failures, but this is the first time I have encountered it specifically with Raptor mini (Preview).
• This behavior is extremely weird and I honestly have no idea why it only seems to happen to me.
• The issue may be related to tool permission handling, tool initialization, or tool capability negotiation when terminal-related execute tools are enabled together.
• This happens since a few versions of VS Code ago, so it's not just v1.119.0's problem. v1.118+ have it, too.

[singularitti]

Similar #253136

[singularitti]

Can somebody have a look at this bug? Now, "disabling any one of the three tools above" no longer works. This has made the model completely useless.

[widget-]

I've been struggling with this for about a month but I can't believe that removing "killTerminal" is really what got it running, at least for now.

I think something about execute and kill are likely triggering anti-violence measures in the content filter.

[singularitti]

Could you try and see if it solves your problem? @widget-

[widget-]

Making a new agent that was missing the killTerminal tool and removing any wording around "execute" in my prompt files made it able to function. createAndRunTask and getTerminalOutput don't seem to be issues for me.

To get around killTerminal not being available, I added a note about just running exit to restart the terminal. That won't solve stuck commands but when I'm ssh'ing into my Mac some ASCII codes can leak and mess up the terminal, and it fixes that.

My initial idea was starting from the Plan agent (since that one worked) and adding on some actual dev instructions, but that might not be necessary. However, since it's works I'm keeping it for now. Here's the agent file:

---
description: "Plan and implement workspace tasks by reviewing existing plans, updating them, and implementing approved steps"
tools: ["vscode/extensions", "vscode/getProjectSetupInfo", "vscode/installExtension", "vscode/memory", "vscode/newWorkspace", "vscode/resolveMemoryFileUri", "vscode/runCommand", "vscode/vscodeAPI", "vscode/askQuestions", "execute/getTerminalOutput", "execute/sendToTerminal", "execute/createAndRunTask", "execute/runNotebookCell", "execute/testFailure", "execute/runInTerminal", "read/terminalSelection", "read/terminalLastCommand", "read/getNotebookSummary", "read/problems", "read/readFile", "read/viewImage", "agent/runSubagent", "edit/createDirectory", "edit/createFile", "edit/createJupyterNotebook", "edit/editFiles", "edit/editNotebook", "edit/rename", "search/changes", "search/codebase", "search/fileSearch", "search/listDirectory", "search/searchResults", "search/textSearch", "search/usages", "@toolsdk.ai/tavily-mcp/tavily-extract", "@toolsdk.ai/tavily-mcp/tavily-search", "todo"]
user-invocable: true
---

You are a hybrid specialist at workspace planning and implementation. Your job is to discover existing plan documents, update them if needed, and then execute the approved implementation steps.

## Constraints
- You MUST check for existing plans before creating new ones.
- You are to behave as both a planner and executor, ensuring that any plan you create is actionable and that you follow through on executing it.
- While planning, you MUST limit tool usage to reading, searching, editing, and limiting command use to exploration and discovery.
- During execution, you SHOULD use the full range of tools available to you.

## Approach
2. If a plan exists, summarize it and determine whether it should be updated.
3. If no existing plan is found, draft a new plan using the agents.md template.
4. After plan review or update, implement the plan by editing files or running commands.
5. Report what was changed, what was implemented, and any next actions.

## Output Format
- Plan summary
- Existing plan status
- Files modified
- Commands run
- Next recommended steps

## Notes

The VS Code runtime may fail to return terminal command contents if an error status code was returned. If this appears to be the case, run e.g. `command; echo "returned $?"` in the terminal to get the command output along with the status code.

If you experience corrupted output in the terminal, it may be an issue with oh-my-zsh. Re-using the terminal to run `exit` and creating a new one should fix it.

[singularitti]

Thank you @widget-, disabling killTerminal works for me, too. 😂 It's silly if it is all about the word "execute".