bmad-review-edge-case-hunter

Provide the diff as input and the review walks every branching path reachable from the changed lines, reporting only boundaries lacking an explicit guard. Findings come back as a JSON array with location, trigger condition, a guard snippet, and the potential consequence.

It derives edge classes from the content itself rather than a fixed checklist. Typical findings include missing else or default branches, unguarded null or empty inputs, off-by-one loops, arithmetic overflow, implicit type coercion, race conditions, and timeout gaps.

Yes. When no diff is provided, the entire file or function becomes the analysis scope and every branch within it is walked. The rest of the codebase is ignored unless the provided content explicitly references external functions.

Yes, a secondary deletion check runs when a diff removes meaningful code. It flags removed behavior or contracts that were neither re-established nor intentionally retired, with a confidence rating of high, medium, or low.

Edge case review is method-driven: it mechanically enumerates paths and reports only unhandled conditions, without judging code quality. Adversarial review is attitude-driven, hunting for weaknesses broadly, so the two approaches are orthogonal and complementary.