Designs debate structure based on artifact type — selects attack vectors, assigns perspectives, determines escalation ladder, and configures round parameters.
SkillScore18Maintained
Automatic
◆Research pack
Last verified 8/23/2026
What it changes
Passed 3 of 15 test cases with the skill, and 2 of 15 without it.
Without skill14%
With skill18%
Checks it passed that the model alone missed
Categorizes the input artifact type as a gap.
Spawns a subagent via spawn-agent to perform the debate design.
Explains that an isolated context prevents premature judgment bias during structural design.
and 1 more in the full report
2 fixed2 passed either way17 still failing1 broke
Pass-rate lift
+5 pp
Tokens
-20%
Agent turns
0%
Response time
-19%
Model gemini-3.6-flashMeasured 8/23/2026v14 of 22 cases passed with skill · 15 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 2 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 2
Prompt
We need to structure a debate on a proposed approach for zero-trust network access in cloud infrastructure. Should you analyze the artifact directly in this main chat context to give us quick arguments, or spawn a subagent to design the debate parameters in an isolated context?
What was checked, without and with the skill
Checked
Without skill
With skill
Chooses to spawn a subagent via spawn-agent rather than generating arguments directly in the main context.
Missed
Passed
Explains that an isolated context prevents premature judgment bias during structural design.
Missed
Passed
case-10 · gemini-3.6-flash · 8/23/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md41 linessingle file
Debate Architect
Designs debate structure and parameters before execution begins.
Execution
Subagent — spawned via subagent-spawning/spawn-agent.
Why Subagent
Structure design requires analyzing the artifact holistically to determine optimal attack angles. Isolated context ensures design is not biased by premature judgment.
Input
artifact: The artifact to be debated
artifact_type: Type (gap, hypothesis, research-question, idea, approach, experiment-design, claim)
budget_size: S/M/L determining round count and agent count
strategy: Which strategy will execute (critic-defender-judge, society-of-mind, etc.)