emerge/challenge-assumptions
Adversarial reviewer personality for architecture discussions. Use when a user requests a design review, architecture review, system design critique, tech stack decision, RFC review, or devil's advocate perspective on trade-offs. Makes Claude challenge assumptions instead of agreeing — questioning scalability assumptions, identifying single points of failure, challenging technology choices, and probing for edge cases rather than validating decisions.
97
100%
Does it follow best practices?
Impact
94%
1.25xAverage score across 5 eval scenarios
Passed
No known issues
Evaluation results
95%
39%Review a Technology Decision for a Financial Transactions System
Technology Proposal Challenge
Does not accept the technology
100%
100%
Names a specific alternative
100%
100%
Asks 2-year scenario question
48%
88%
Scenario is specific
25%
90%
Question is direct
0%
100%
100%
25%Review an RFC for a Healthcare Appointment Booking System
Vague Requirements Quantification
Does not assume availability target
100%
100%
Asks for specific availability number
100%
100%
99.9% vs 99.99% distinction
0%
100%
Does not assume performance target
100%
100%
No architecture proposed yet
100%
100%
100%
20%Review an Architecture Proposal for a Small Internal Tool
Over-Engineering Detection
Directly flags over-engineering
100%
100%
Names specific component
100%
100%
Suggests simpler alternative
100%
100%
Asks to be convinced
0%
100%
Does not validate the design
100%
100%
79%
12%Review a Payment API Design
Under-Engineering and Vague Approval Push Back
Rejects vague approval
100%
100%
Asks about weakest part
46%
53%
Asks about fallback
70%
20%
Flags missing auth/security
100%
100%
Flags missing monitoring
0%
90%
Production failure scenario
86%
93%
Cost of fixing later
0%
60%
Does not approve the design
100%
100%
100%
Deep-Dive Architecture Review for an E-Commerce Order Platform
Phase 2 High Tension Deep Dive
Challenges scalability assumption
100%
100%
Identifies single point of failure
100%
100%
Challenges a dependency
100%
100%
Technology choice challenged
100%
100%
Specific probing questions
100%
100%
Does not approve overall
100%
100%
- Evaluated
- Agent
- Claude Code
- Model
- Claude Sonnet 4.6