If a turn is cancelled while the context compressor is waiting on its LLM summary, Agent.chat() rejects before it reaches the existing cancellation handler. The new user message remains as a bare trailing user message, so a later request can treat the cancelled task as unfinished.
What changed
Move the initial user-message append and maybeCompress() call inside the existing try block.
Reuse the current [interrupted by user] history marker for cancellation during compression.
Add a regression test that cancels a summary request and checks the resulting history.
Verification
npm test — 55 passed, 1 skipped (the existing platform-specific SIGKILL test)
git diff --check
The failure was reproduced locally with a cancellable LLM stub before the fix. This change was AI-assisted and reviewed against the current min branch.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What breaks
If a turn is cancelled while the context compressor is waiting on its LLM summary, Agent.chat() rejects before it reaches the existing cancellation handler. The new user message remains as a bare trailing user message, so a later request can treat the cancelled task as unfinished.
What changed
Verification
The failure was reproduced locally with a cancellable LLM stub before the fix. This change was AI-assisted and reviewed against the current min branch.