code-review

Provide a diff source such as a branch, commit range, staged changes, or pasted diff, and the workflow launches parallel review layers. Each layer examines the diff independently, then findings are triaged into patch, defer, decision-needed, or dismiss categories.

You can review uncommitted changes, staged changes only, a branch diff against a base branch, a specific commit range, a user-provided diff, or a list of files. The workflow validates each source and halts if the diff is empty or unparseable.

Yes, it runs in no-spec mode where the Acceptance Auditor layer is skipped. Findings that would require human decisions are reclassified as patch or defer since there is no spec to resolve ambiguity against.

The failed layer name is recorded and the review proceeds with findings from the remaining layers. If zero findings remain and any layer failed, the workflow warns that the review may be incomplete instead of declaring a clean review.

Yes, when a story key is known, the workflow updates the story file status to done or in-progress based on review outcome and syncs the matching entry in sprint-status.yaml while preserving file structure and comments.

Very large diffs over roughly 3000 lines should be chunked by file group to stay effective. Also, decision-needed findings require a spec file, and without subagent support you must run each reviewer prompt in separate sessions manually.