Repository navigation
test: Claude Agent compaction sends its summary request as the prompt - #1797
Merged
Merged
Conversation
Compaction runs on a fresh tools-off Claude Code process. Cut mid tool loop, its history ends on tool results (or a call) followed by the summary request. freshContent used to fold the request into the text checkpoint and add "Continue from where the checkpoint ends.", so the model carried on with the task or asked the user a question, and that reply was installed as the summary. #1792's change to freshContent fixed this; this pins both shapes.
Merged
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Claude Agent compaction could install a reply instead of a summary.
Compaction runs on a fresh, tools-off Claude Code process. When it is cut mid tool loop, its history ends on tool results (or a call) followed by the summary request.
freshContenttook the trailing user messages as the prompt only when none held atool_result, so the request went into the text checkpoint, followed by "Continue from where the checkpoint ends." In a long conversation the model carried on with the task, or asked the user a question, and that text was installed as the compaction summary.#1792's change to
freshContent(the prompt is what follows the lasttool_result) fixed this too. This PR adds a regression test for both shapes: history ending on tool results, and on a call.