---
name: debate-critic
description: Generates structured criticism from attack stance using Toulmin model. Produces claims, grounds, warrants, and rebuttals targeting artifact weaknesses.
execution: subagent
prompt: ./prompt.md
input: artifact (string), escalation_level (string), attack_vectors (string)
used-by: [multiagent-debate]
---
# Debate Critic
Generates structured adversarial criticism targeting artifact weaknesses.
## Execution
Subagent — spawned via subagent-spawning/spawn-agent.
## Why Subagent
Attack generation requires dedicated adversarial stance without contamination from defensive reasoning. Isolated context ensures pure opposition.
## Input
- **artifact**: The artifact being debated (full text)
- **escalation_level**: Current level (L1-surface, L2-structural, L3-foundational)
- **attack_vectors**: Specific angles to attack (from debate-architect)
## Output
- **attacks**: List of structured attacks (claim + ground + warrant + rebuttal anticipation)
- **confidence**: How confident the critic is in each attack (0.0–1.0)
- **escalation_suggestion**: Whether to escalate, maintain, or de-escalate
## Budget
One unit = one set of attacks for one round.