| name | idea |
| description | Validate and plan one net-new idea, greenfield capability, or immature product surface. Uses research-first evidence, a real kill criterion, and explicit GO before implementation. |
Idea
Thin entry to plan with IDEA pinned and research=hybrid by default. Do not duplicate planning spine here.
- Set goal anchor and retrieve memory before research.
- State proposition, proof-user, existing alternative, adoption evidence, and kill criterion.
- Research only decision-changing unknowns. Inspect buy, reuse, integrate, and fork candidates before proposing custom build.
- Produce validation verdict:
BUILD, DON'T-BUILD, or NOT-NOW, with evidence for, evidence against, and kill-criterion result.
- For
BUILD, produce PRD/build context, then run plan shared spine: approaches, dependency graph, parallel batches, spec-drift, delete/merge, adversarial review, correctness gates, and explicit GO.
If substantial relevant implementation already exists, state reclassification and run plan as CHANGE. DON'T-BUILD and NOT-NOW are complete outcomes; no implementation follows them.