planlisted

# plan One axis: **which artifact you want**. The stages form a maturity ladder, and you enter at the rung where your uncertainty actually sits — running an earlier one when you already have its output is the most common waste here. | you want | invoke | reference | it returns | |---|---|---|---| | the space of options | `--stage ideas` | `references/ideas.md` | an idea board from alternating discussion and research rounds. Generative, **not** ranked | | one option out of several | `--stage shortlist` | `references/shortlist.md` | a ranked shortlist plus hybrids: persona ideators diverge, a red team kills the weak by majority, champions steelman the survivors | | to know exactly what is asked | `--stage spec` | `references/spec.md` | a spec, reached by **one question per round** with an ambiguity gate, topology-first scoping and an approval gate before handoff | | a plan | *(no stage)* | `references/plan.md` | one plan: N planners draft under different lenses, judges score them **against each other**, a synth step merges the winner with the best of the rest | | a plan you can bet on | `+ --deep` | `references/deep.md` | the same tournament, then a sequential Planner→Architect→Critic pass. Auto-widens to a pre-mortem and expanded test plan on auth, migration, destructive, PII or API-break signals | `--deep` is the depth axis, spelled the same as in `review --deep`: same task, more instances, an adversarial or sequential check on top. ## Which rung - Nothing written down y