Agent
peer-critic
Objects as another designer would
@cf/openai/gpt-oss-120b
With image input: @cf/meta/llama-4-scout-17b-16e-instruct
Five critics predict the objections your work will draw, each arguing only from what they are accountable for and none of them seeing the others. Then the answers get built and labeled - defense, concession, or trade - and the questions you should ask get written. Walking into a crit without this means improvising, which is how good work gets cut.
Execution
- peer-critic + engineering-critic + business-critic + user-advocate + self-critic · independent parallel reviews
- defense-builder
- question-crafter
- sheet-compiler
The team
Agent
engineering-critic
Objects as the person who has to build it
@cf/qwen/qwen2.5-coder-32b-instruct
With image input: @cf/meta/llama-4-scout-17b-16e-instruct
Agent
business-critic
Objects as the person accountable for the outcome
@cf/openai/gpt-oss-120b
With image input: @cf/meta/llama-4-scout-17b-16e-instruct
Agent
user-advocate
Objects on behalf of the person who has to use it
@cf/openai/gpt-oss-120b
With image input: @cf/meta/llama-4-scout-17b-16e-instruct
Agent
self-critic
Names what the maker already knows is weak
@cf/openai/gpt-oss-120b
With image input: @cf/meta/llama-4-scout-17b-16e-instruct
Agent
defense-builder
Merges the five critics and labels every answer
@cf/openai/gpt-oss-120b
Agent
question-crafter
Writes the questions to ask the room
@cf/openai/gpt-oss-120b
Agent
sheet-compiler
Compiles the one-page crit sheet
@cf/openai/gpt-oss-120b