What problem does it solve? Before real users can be interviewed, teams need early signal on whether a brief, spec, prototype, or running app will hold up. This Skill builds panels of synthetic personas from existing research material, interviews them under decision pressure, walks flows through their eyes, and reports only findings that repeat across runs — always labeled as hypotheses, never as validation. ## Core Features & Use Cases - Evidence-grounded persona panels: Personas are built only from sourced material (interviews, data, documents), with no demographics and every line tagged by evidence tier; assumption-heavy personas are labeled as such. - Interviews under pressure with repeated runs: Past-tense questions, simulated decisions under budget and time pressure, fresh panels per run, and findings reported only when they survive every run. - Three stances and parity checks: Validate an existing product, simulate unreachable client users, or adversarially attack your own idea; when real interview notes exist, a held-out parity check measures where the panel diverges from real people. - Use Case: You have a product brief with risky assumptions but no access to users yet. Run the skill on the brief to get repeated barriers, contradictions, and a concrete interview plan of what to confirm with real users. ## Quick Start Ask the agent to run the synthetic users skill on your product brief for the onboarding flow, for example: "Run om-synthetic-users on product-brief.md with the onboarding flow and adversary stance."
