workforce

Delegation policy and model rankings for subagents: which model (gpt-5.6-sol / opus-5 / fable-5.1) to pick for which task, how to be an orchestrator instead of burning your own context, and how to run gpt-5.6-sol via codex-wrapper. Only use this skill if you are fable-5.1 model and you are the main agent. Do not use it if you are Opus/Sonnet/subagent etc.