Fix skipped tool response contamination with success message #275960
+90
−3
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
When a tool call is skipped (either pre-execution or post-execution), the pastTenseMessage from prepareToolInvocation was still being included in the response sent to the language model. This caused confusing messages like "Here is some relevant context from the web page..." to appear alongside the skip message. The fix sets toolResultMessage to an empty string when a tool is skipped, which overrides the pastTenseMessage property on the ChatToolInvocation object. This ensures only the skip message is sent to the language model. Changes: - Set toolResultMessage to empty string when tool is skipped (pre-execution) - Set toolResultMessage to empty string when tool results are skipped (post-execution) - Added tests to verify the fix Co-authored-by: dmitrivMS <9581278+dmitrivMS@users.noreply.github.com>
Copilot
AI
changed the title
[WIP] Fix bogus text in skipped fetch tool response
Fix skipped tool response contamination with success message
Nov 6, 2025
Copilot finished work on behalf of
dmitrivMS
November 6, 2025 23:49
Contributor
|
@copilot This does not fix the problem, try again. |
The previous fix didn't work because JavaScript/TypeScript treats empty strings as falsy. The condition `if (result?.toolResultMessage)` would be false when toolResultMessage is '', so the pastTenseMessage never got overridden. Changed the condition to `if (result?.toolResultMessage !== undefined)` to properly handle empty strings and ensure the pastTenseMessage is cleared when tools are skipped. Co-authored-by: dmitrivMS <9581278+dmitrivMS@users.noreply.github.com>
Contributor
Author
Fixed in 97c47e4. The previous approach didn't work because empty string |
Copilot finished work on behalf of
dmitrivMS
November 7, 2025 07:52
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Skipped tool invocations were leaking success messages into the language model context. When a user skipped a tool approval (e.g., fetch request), the Chat Debug view showed both the preparation success message and the skip message:
Changes
The
pastTenseMessagefromprepareToolInvocationpersisted onChatToolInvocationeven after skip. The initial fix settoolResultMessage: ''in the skip result, but this didn't work because the conditionif (result?.toolResultMessage)treated empty strings as falsy in JavaScript/TypeScript.Fixed by changing the condition to use strict undefined checking instead of truthiness checking.
languageModelToolsService.ts
toolResultMessage: ''to tool resulttoolResultMessage: ''to tool resultchatToolInvocation.ts
if (result?.toolResultMessage)toif (result?.toolResultMessage !== undefined)to properly handle empty string valueslanguageModelToolsService.test.ts
Fixes contaminated language model context when tools are skipped.
Original prompt
💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.