review-plan / default

Reviewing an AI action plan for safety requires validating required structure, safety constraints, and context-faithfulness. This involves reading the action index and latest plan stage to verify seven required sections, including goals, tools, test plans, and rollback procedures.

Context reference integrity in AI plan review confirms that captured context_refs are explicitly honored across the plan or formally dismissed with reasoning. Silent omission of context_refs triggers an automatic rejection to prevent execution drift.

Validating rollback safety and testability requires checking the action plan for explicit success criteria, a dedicated test plan section, and a comprehensive rollback strategy. The plan must demonstrate reversibility before passing the approval gate.

Yes, playbook-other requires an explicit artifact declaration for plan approval. Plans within this workflow must stand alone and declare artifacts explicitly; missing declarations or context_refs issues result in section-referenced rejection feedback.

When an AI plan fails the approval gate review, the system rejects it with a detailed section-referenced issue list. This rejection highlights specific failures in artifact declaration, context honoring, or missing structural sections.

You need a strict safety-first plan review for AI workflows when an action's latest plan stage must stand alone and be validated against captured context. It prevents unsafe execution by enforcing testability, reversibility, and governance constraints.