Run multi-model council debate engine

council-engineworkflowsetup L30
AAARRRCCC/claude-skills

Causal-lift measurements

debate-synthesis+2pp vs no-skill baselinewith-skill 36% · baseline 34%

Measured by running the task with and without this artifact, K=5, graded by deterministic checks — no LLM judging.

What it does

Engine behind the /council skill: a multi-model, multi-persona council that deba

Best for

Multi-model, multi-persona analysis of complex questions with synthesized verdicts.

Inputs
  • · args: string or object
Outputs
  • · verdict JSON
Requires
  • · test runner
Failure modes
  • · Iteration limit reached
  • · Verification gates fail
  • · Multi-implementation divergence not resolved
Trust signals
  • · Explicit phase orchestration with gates
  • · Remediation/iteration loops with convergence criteria