simulate-audience

When to use

When "would target users respond positively" is the ground truth:

• Ad copy, cold email, social post — does the hook land?
• Landing page, pricing page — would visitors convert?
• Logo, brand mark, social image — does it feel right for the business?
• Blog / video script — useful or generic?
• Onboarding flow — confusing or clear?

Don't use for: code correctness, API contracts, data integrity — those go to Playwright / gstack /review / verify-criteria.

What you do

1. Design the panel yourself

Read the task + artifact + business context. Design 4-6 personas FOR THIS BUSINESS — not generic demographics. Use JTBD (jobs-to-be-done) framing: who they are, what they're trying to get done, what turns them off vs on.

One persona MUST be a hardened skeptic. Someone who's seen every pitch, notices AI-slop / stock / templates immediately, distrusts hype. This forces critique — without it panels collapse into "everyone loves it" mush.

2. Have each react in character

For each persona, in their voice, react to the artifact. Ask whatever follow-ups matter — don't follow a fixed question list. A cold-email persona will have different questions than a logo-design persona.

3. Synthesize across the panel

• Themes — concerns or praises that ≥ 2 personas raise
• Dealbreakers — issues serious enough to block ship (≥ 3 personas, or
any single trust/safety concern)
• What to fix, ranked

Output

Return JSON on your final message:

{
  "personas": [
    {"name": "<name + 1-line identity>", "why_in_panel": "<their angle>"}
  ],
  "reactions": [
    {"persona": "<name>", "reaction": "<2-3 sentences in character>",
     "score_if_relevant": 1-10}
  ],
  "themes": ["<recurring concern or praise>"],
  "dealbreakers": ["<ship-blocker or empty>"],
  "verdict": "ship it" | "needs work" | "do not ship",
  "what_to_fix": ["<prioritized concrete change>"],
  "one_line": "<warm non-technical summary for the creator>"
}

What NOT to do

• Don't use generic demographic personas — they give generic feedback.
Make them specific to THIS business.
• Don't skip the skeptic. Panels without critique produce mush.
• Don't pretend synthetic personas replace real user testing. For actual
shipping decisions at scale, 5 real users on Prolific ($30) beats 50 synthetic personas. This skill is an iteration-speed tool.
• Don't impose a rigid question list. Adapt to the artifact.