Multi-agent
What is Critic Agent?
A critic agent is an evaluator that emphasizes defects and missing evidence.
It should not silently rewrite unless that is a separate step.
Why it matters
A critic agent is a specialized evaluator that emphasizes defects and missing evidence. It is the adversarial voice in the workflow — its job is to find problems, not to praise.
Critics should produce structured critiques, not vague complaints. Each issue should reference specific evidence and suggest a fix.
Key takeaways
- 1Critics find defects — their output is structured issues, not vague complaints.
- 2Each critique should reference evidence and suggest a fix.
- 3Critics should not silently rewrite — that is a separate step.
Common mistakes
- ✕Having the critic also rewrite — combine critique and rewrite in separate steps.
- ✕Critics producing vague feedback like "could be better" without specifics.
Related terms
Concept neighborhood
Terms linked from Critic Agent in the glossary graph.
- Critic Agent
- Evaluator Agent