crew-plan

A stress-tested implementation plan is generated by iteratively drafting, self-critiquing, and revising your task across multiple rounds to make assumptions, risks, and verification steps explicit. This multi-pass process reduces hidden gaps before execution begins.

Multi-pass plan convergence is a technique that refines an implementation roadmap through up to three iterative drafting and critique rounds. You need it for software builds or design-to-implementation scenarios requiring explicit risk management and safer execution.

You add risk management and verification steps by generating an implementation plan that mandates explicit assumptions, identified risks, and verification criteria for each decision. The plan is self-critiqued and revised until these safety checks are fully defined.

Yes, you can use multi-agent critique by requesting a bounded skeptic or reviewer pass during planning. If subagents are available, this adds an automated debate layer to further stress-test your implementation plan before execution.

Yes, crew-plan saves accepted implementation plans to `.codex/crew/seed-<timestamp>.md` and updates `.codex/crew/seed-latest.md`. This ensures your stress-tested roadmaps are stored as seeded artifacts for future reference and traceability.

Automated multi-round planning is limited by its dependency on available subagents for multi-agent critique delegation, and it requires iterative drafting time. It focuses on producing safer execution plans rather than executing the implementation itself.