Repository navigation
Conversation
… the other part's file marked
There was a problem hiding this comment.
🤖 NCode Review at db53af5: 4 new findings on the changed lines.
|
Superseded by the review at 738aa16. |
|
/review |
|
Autopilot: ready to merge as far as this run can tell. One review round: the CodeQL false positive and two nits were fixed (f034e7d, ee8a6d7), and the re-review at ee8a6d7 raised only a repeat of the changelog nit already answered. No blocker or should-fix is open. Follow-ups filed after the merge: the judge-spend test (needs the sum moved out of |
|
/review |
🤖 Review completeVerdict: ✅ Approve — Nothing in this pull request's own changes differs from the last review at ee8a6d7, so nothing was reviewed again. No further findings above nits. 0 reviewers run — 0 findings (run details)
Trial: findings are posted as review threads and in the check run. $0.00 on Amazon Bedrock, 4s, at 738aa16. (how to interact) 📋 OutcomesMerged at 738aa16: 1 fixed, 3 dismissed, 0 acknowledged, 0 ignored.
|
Adds the tests declined past the review round limit on #64, #68, #69 and #71. A new
test/review-prompts.test.tschecks, throughreview()with a stub agent, that each part's reviewers get that part's diff with the other part's files marked, that each reviewer call gets its part's rules and its own charter, and that project reviewers run on their configured tier.test/fallback.test.tsgeneralizes its helper intoreviewOnFakeSdk()and tests the Bedrock rerun of the selection and context calls, pluswithoutToolsdirectly, through the fake Agent SDK.Only test files and
CHANGELOG.mdchange; the judge-spend (#73) andanswer()(#75) tests need a non-test refactor first and are left as follow-ups.Closes #96