playbooks/radical-product.md

# Radical Product Engine

Use only for requests such as radical AI product ideas, 10x, eliminate UI, replace a workflow, autonomous product, or Vibe Build. Do not turn a minor optimization into a radical claim.

1. Inspect the app, repository, actual signals, and current workflow. Separate observed facts from assumptions.
2. Identify the interaction to eliminate, not a feature to add. Challenge data gathering, handoffs, status checks, copying context, and routine choices.
3. Produce exactly five AI-central paradigms. Each must replace a complete interaction, state a 10x hypothesis, data required, trust boundary, moat, MVP, metric, and kill condition.
4. Rank by status-quo break, 10x potential, MVP feasibility, defensibility, and evidence. Select one only with an explicit rationale.
5. Design the replacement flow: minimum input, bounded action plan, material-decision review, exception path, undo/rollback, and outcome evidence. Prefer no dashboard.
6. Build the MVP through Observe → Suggest → Approve → Bounded Autopilot. Autonomy is restricted to pre-authorized, reversible, low-risk actions.

Use `forgemind radical analyze --goal "<outcome>" --json`, then `radical select`, `radical blueprint`, and `radical shadow-mode`. Use `experience market-case`, `finance`, tester evidence, and `$forgemind-complete` after selecting the bet.

Never invent behavior, demand, ROI, permission, or model capability. Never permit secrets, production access, destructive actions, irreversible changes, external spend, or high-stakes decisions without explicit approval.