/agent-evaluate
Multi-Agent Evaluate
Parallel persona agents score content independently
Bespoke TierContent & Editorial
Overview
Runs 5-7 persona agents in parallel against a content corpus, then calculates consensus scores using a two-track formula. Detects tensions between agents and maps content to output sections. Each agent applies critical thinking, not field-checking.
What It Does
- Spawns parallel persona agents with distinct evaluation criteria
- Calculates two-track consensus scores (editorial-picked vs. agent-only)
- Detects tensions between agents and boosts scores for contested items
- Maps evaluated items to output sections based on agent consensus
Inputs
- Content corpus to evaluate
- Persona agent configurations
- Editorial picks
Outputs
- Agent evaluations with scores
- Consensus rankings
- Tension map
- Section pre-mapping
Example
/agent-evaluate
Run evaluation on 35 classified articles. The Builder and Revenue Leader both STRONG_PICK an architecture teardown while the Contrarian flags it as overhyped. Tension detected, score boosted, flagged for the debate phase.
Deep Dives
Ready to use /agent-evaluate?
This skill ships with every Knowledge OS installation. Set up your system in 90 minutes.
Built and maintained by Victor Sowers at STEEPWORKS