What problem does it solve? Coordinating work between an orchestrating AI and an external coding model requires consistent rules for model selection, dispatch, retries, and failure handling; this Skill defines that contract so implementation tasks are delegated to Codex GPT-5.6 predictably and transparently. ## Core Features & Use Cases - Tiered Model Routing: Classifies each task as mechanical, standard, or complex and routes it to gpt-5.6-luna, gpt-5.6-terra, or gpt-5.6-sol with matching effort levels. - Preflight & Health Checks: Verifies the Codex CLI is installed, authenticated, and new enough to support the GPT-5.6 tiering models before the first delegation. - Fix Cycles & Fallback Ladder: Resumes Codex sessions with delta-only fix instructions after review findings, and falls back to a Claude implementer subagent after repeated failures, always announced to the user. - Use Case: While executing a multi-task implementation plan, the orchestrator dispatches each coding task to the appropriate Codex tier, reviews the returned diff, sends back only the review findings for fixes, and records the model and fix-cycle count in a progress ledger. ## Quick Start Ask Claude to implement the next task in the plan and have it delegated to the appropriate Codex GPT-5.6 model tier.
