Repository navigation
self-review: report adjacent issues only when found - #3
Merged
Merged
Conversation
The skill required an "Adjacent issues: none found" line in every summary, which added noise to almost every hand-off. Agents now add the line only when the ±20-line scan finds something, and step 8 keeps clean checks out of the summary. Tested with single-shot Sonnet runs: summaries of a clean change mentioned the scan 5/5 times before and 0/10 after. Planted adjacent issues were still named 10/10. Fixes #2
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #2.
self-reviewtold agents to end every summary withAdjacent issues: none found in ±20 lines of touched hunks.The idea was that an agent that skipped the scan would have nothing to write, but an agent can write "none found" without looking, so the line was mostly noise.Agents now add the
Adjacent issues:line only when the scan finds something. The scan itself, the six patterns, and the test-fixture carve-out are unchanged.Removing the line wasn't enough on its own. In testing, agents swapped it for a sentence like "I checked the surrounding lines and found nothing unsafe." Step 8 now says what the summary contains and that checks that came up clean aren't part of it. A side effect is that summaries also stop listing other checks that passed, like "no new logging, no warnings."
Testing
Single-shot Sonnet runs, with the skill as system prompt and a finished change to hand off. 5 runs per cell unless noted.
The no-skill runs caught every planted issue too, so these tests show the new wording cuts the noise without suppressing findings. They don't show whether the skill makes agents scan more. A
97-benchrun would answer that.npm testpasses.