What problem does it solve? After a feature is implemented, teams need a disciplined, read-only review that compares the work against a fixed baseline, tracks findings across multiple rounds, and hands off approved fixes without letting the reviewer accidentally modify code. ## Core Features & Use Cases - Baseline-anchored review: Reads the immutable base_sha from .planning/executions/<feature>/forge-state.json and reviews committed, staged, unstaged, and untracked changes since that point. - Iterative findings tracking: Maintains a durable report at .planning/reviews/<feature>.md with stable finding IDs (REV-001, ...), severity ordering, and verdicts (CHANGES_REQUIRED / OK). - Apply handoff workflow: A separate tab applies only user-approved findings and records changed files plus narrow test results in the same report. - Use Case: After forging a new checkout feature, invoke the review to get a prioritized list of blockers and regressions, approve two findings, and have a second tab apply and verify them while the report tracks resolution across rounds. ## Quick Start Ask the assistant to review the recently forged feature using the saved execution state and write the findings report.
