journey-simulationlisted

# journey-simulation ## Overview Spawn isolated subagent to **execute a real task** under identity-as-fact briefing + sandbox spec. Subagent never knows it's observed. Gateway extracts funnel post-hoc from narration. **Persona is a distribution, not a single person.** Persona-slug names a group; subagent samples one path. Read each waypoint as "how thinned is this group?" — not "did this individual finish?" Funnel attrition is the unit of signal. ## When to use - Caller asks to "simulate a user", "test our flow", "will users get stuck on X", "cognitive walkthrough", "cold-eyes review" - Pressure-testing a flow before real-user testing or after a redesign - Reviewing whether a doc / README / game loop / future-self artifact is comprehensible cold **Skip when:** caller wants ground-truth real-user data (use real-user testing); caller wants design recommendations (this skill observes, doesn't prescribe). ## Phase 0 — Clarify & de-bias Run before dispatch. Caller-in-loop. **Never skipped.** 1. **Ask:** what artifact, what flow, desired outcome of this test, your concern / hypothesis? 2. **Surface caller's bias:** "you suspect X — the sim must not be primed to find X. Restate inputs in neutral terms." 3. **Ask:** realistic personas (not aspirational ICP)? Skeptic / return / negative cases worth running? 4. **Curate:** strip project jargon, success-bias, design rationale, vision-statement leakage. 5. **Structural-cue check:** if curated persona reads as constructed (over-p