playbooks/leap.md

# Leap Autonomous Disruption Contract

## Outcome

Produce a tested, reversible MVP from a new idea or existing app—not an idea list. The first output is the decision record; the next action is implementation.

## Sequence

1. Inspect the app and classify repository facts separately from user, market, and competitor assumptions.
2. Generate five AI-central interaction replacements. Each must remove a complete workflow or deliver an order-of-magnitude step reduction.
3. Score the candidates for disruption, AI centrality, feasibility, evidence strength, moat, monetization, and reversible trust fit.
4. Select the strongest reversible bet automatically and retain a credible contrarian. Create its market-chance, business-case, kill-condition, rollback, shadow-mode, and completion contracts.
5. Continue in Complete with YOLO speed only for the selected bounded MVP. Build all directly affected states and callers, then test the decisive task.
6. Finish with verification, GUI and accessibility evidence where relevant, risks, rollback readiness, residual gaps, and a release decision.

## Non-negotiable evidence rules

- A repository can show technical context; it cannot prove demand, pricing, customer behavior, or ROI.
- Preserve the market and finance model as assumptions until qualified evidence is attached.
- A score ranks alternatives; it does not turn an option into a product fact.
- Never report a test, customer validation, or business result that has not been recorded.

## Autonomy boundary

Proceed without routine questions. Pause only for secrets or credentials, production access, data deletion, irreversible migration, external spend, or a high-stakes decision. At such a boundary, finish every independent action and request the smallest possible decision.