loamconfiguring-agentslisted

# loam::configuring-agents ## Overview Run a structured debate, conference, or deliberation between agents so they reach consensus on a goal. The skill produces a prepared debate plan — hub prompt, distinct role briefs, convergence prompt, round count, stopping rule, forcing-field deliverable, and the exact first hcom action — and presents it for human approval. No agent contact happens before approval. If approval is denied or absent, the prepared plan is the deliverable and the skill stops. Core principle: produce a runnable consensus protocol, not just advice. This skill exists to prevent five baseline failures: - stopping at the parameter interview when safe defaults would compose a complete prepared plan - firing any hcom send, spawn, thread seed, or launch before explicit per-debate human approval - drifting into project implementation instead of staying in debate/consensus scope - emitting launch guidance with invalid OpenCode spawn flags or ambiguous bare model names - treating a one-off debate as a reusable saved team config — that contract is retired ## When To Use Use this skill when the user wants any of the following: - a structured debate between agents arguing distinct positions - a conference or roundtable to work out a goal - deliberation that converges on a single deliverable - agents to reach consensus on a decision, design, or direction - an adversarial review where two or more positions must argue and converge Do not use this skill for: - produci