divergelisted

# /diverge — Expand the option space before commit Use `/diverge <task>` when the task admits multiple credible framings and you want to see them side-by-side with explicit tradeoffs *before* a planner commits to one. ## When to use - Open-ended prompts: *"how should we approach the rate-limit handling?"*, *"what's the best architecture for this?"*, *"is there a better way to do X?"* - Architecturally significant choices where the paradigm decision is non-obvious (event-sourced vs CRUD, push vs pull, sync vs async, monolith vs services, library-build vs use-existing). - The model has stated an interpretation and you want to verify it covers the space before commit. - A council surfaced findings that disagree on framing — the lenses are pointing at different shapes of solution. ## When NOT to use - Single-line bug fixes, well-scoped refactors, or any task with an obvious dominant approach. Adding `/diverge` to routine work is friction, not value. - Tasks already mid-execution. `/diverge` is upstream-only — once a planner has committed to a paradigm, use `metis` to stress-test it or `abstraction-critic` to question the fit. - Debugging. Use `oracle` for "what's the root cause?" and `metis` for "what's wrong with this plan?". ## How it works The skill dispatches the `divergent-framer` agent with your task. The agent emits a structured menu: - **3-5 framings**, each with: a 2-4 word name, the mental model in one sentence, what the paradigm makes EASY (1-2 concrete afforda