vibe-strategylisted

# vibe-strategy — direction loop The strategy flow shapes project direction. It reads lessons, brainstorms, and writes the **root** `.spec/` docs. It never writes source code. States: `strategy.brainstorm → strategy.spec → (strategy.compound | idle)`. ## Procedure 1. **Locate.** Read `.agents/flow/state.json`. If not already in a `strategy.*` state, transition in: `bash .agents/flow/scripts/set-state.sh strategy.brainstorm`. 2. **Read lessons first.** On entering `strategy.brainstorm`, read `.spec/lessons.md` so past mistakes shape direction (retrieval, not just recording). 3. **Brainstorm.** Delegate to `superpowers:brainstorming` for the human dialogue. Scratch only — no spec writes yet. When direction is clear, suggest: "direction looks set — move to strategy.spec?" and transition on confirmation. 4. **Spec.** In `strategy.spec`, delegate to the `spec` skill. Write **only** the root docs: `.spec/product.md`, `.spec/tech.md`, `.spec/design.md`, `.spec/plan.md`. Inject those exact paths. Do not touch `lessons.md` or source. Validate with `bash .agents/skills/spec/scripts/validate.sh`. 5. **Compound (conditional).** Most strategy runs surface no durable lesson. Only if one did, suggest `strategy.compound` (delegates to `vibe-compound`); else go to `idle`. Trust the agent to pick — this is "trust the agent" applied to control flow. ## Rules - Caveman **lite** throughout (full sentences, no filler). - Transitions are agent-*suggested*, not