polygnosis - Adversarial multi-model consensus for high-stakes tasks
Runs a multi-model debate pipeline with specialized solvers, critique, formal ranking, synthesis, and quality gating.
Tags
Updated: 2026-09-28multi-agentconsensuspeer-reviewadversarialboardroomreasoningcode-reviewrrfbordareflexionquality-gate
Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Assign specialized personas
- Run parallel model solves
- Perform adversarial critique
- Log severe findings
- Rank solutions formally
- Synthesize ranked solutions
- Check synthesis regressions
- Explain consensus decisions
Inputs
- User objective
- Success criteria
- Model configuration
- Scoring algorithm
- Run artifacts
Outputs
- Final output
- Meta-review
- Run artifact files
- Corrections buffer updates
- Quality gate verdict
Requirements
- Hermes Agent
- Three or more configured models
- Configured PolyGnosis pipeline
- Python runtime
- Background terminal execution
- Completion notifications