bmad-checkpoint-preview

A human-in-the-loop code review for a checkpoint requires a structured walkthrough that combines intent, git diff context, and configuration into an actionable review order. The process guides you through orientation, detail passes, and testing steps.

A risk-focused code walkthrough structures your human review around design intent rather than reading files sequentially. It surfaces high-bite-risk areas like auth, public API, schema, billing, security, and infra changes to guide what to scrutinize.

To validate a git diff during a checkpoint review, the system loads persistent facts and workflow configuration, then generates an actionable review order. This produces orientation, walkthrough, detail pass, and testing steps with path:line callouts.

Yes, the guided review process supports custom workflow configuration. It runs workflow customization resolution and loads persistent facts and language inputs before executing the orientation, walkthrough, and testing guidance steps.

You need a checkpoint review when you must quickly understand what a code change is trying to do and where to focus your attention during human validation, especially when dealing with complex risk areas like schema or security modifications.

An LLM-assisted review for checkpoints requires YAML frontmatter-defined discovery metadata to function. It relies on a generated review trail when available, meaning its guidance is limited by the provided diff context and configuration inputs.