Cook CLI orchestrates Claude Code loops

This feels less like a brand-new AI product and more like a sharp workflow layer for people already living inside agentic coding tools. Its real value is turning ad hoc prompting into repeatable loops, but that same abstraction will only matter to teams who are already comfortable managing sandboxes, gates, and branch races.

• –The `review -> gate -> iterate` pattern is the strongest part; it bakes in quality control instead of treating review as an afterthought.
• –`vN` and `vs` are practical for code generation contests, but they also add complexity that casual users probably won’t want.
• –Supporting Claude Code, Codex, and OpenCode makes Cook more portable than a single-vendor wrapper.
• –The `cook init` scaffolding suggests this is aimed at real project workflows, not just toy prompts.
• –Open-source distribution plus a CLI-first install path should make it easy for power users to adopt, even if the audience stays niche.