ISOLATED AGENT BRAINSTORM PROCESS
Challenge: Increase user engagement for an education SaaS
Goal: 8-10 non-obvious ideas, scored on novelty and viability

FRAME LIBRARY (3 isolated agents, each gets one frame)

Agent 1 - Inversion Frame
'How would we decrease engagement intentionally? What if we made the app intentionally harder to use? Turn each friction point into a feature.'
Output: Require daily commitment + streak counter
Novelty score: 7/10 (habit-building angle novel, but common pattern)

Agent 2 - Adjacent Market Frame
'What if we were building this for elite athletes or musicians, not students? What would they need differently?'
Output: Achievement leaderboards by skill level with coach verification
Novelty score: 9/10 (competitive framing new to education SaaS)

Agent 3 - Constraint Frame
'What if we had zero budget for content, zero ability to hire, one engineer only? What's possible?'
Output: User-generated problem sets with crowdsourced verification
Novelty score: 8/10 (peer review model not in current product)

CRITIC AGENT SCORING
Scores each on: novelty (is this actually new to market), viability (can one engineer build in 3 weeks), business impact (does it move the needle)

SHORTLIST (non-obvious ideas)
1. Competitive leaderboards by skill (score: 8.2/10)
2. Peer problem verification (score: 7.9/10)
3. Habit streak system (score: 7.1/10)

TRAP LIST (ideas that sound good but fail viability)
- AI tutor integration (requires licensing)
- Real-time collaboration (complex infrastructure)
- Mobile-first redesign (timeline conflict)

Generated ideas are genuinely different from single-agent brainstorm.