What problem does it solve? Coordinating multi-agent coding work often hides which model actually plans or executes a task, leading to silent model substitutions, wasted cost, and unverified results. This Skill enforces a transparent plan-execute pipeline in Codex where every delegation is announced, confirmed, and verified before work begins. ## Core Features & Use Cases - Four-Quadrant Model Routing: Assigns planning to Sol/high (upgradable to Sol/xhigh for complex scopes), complex implementation to Terra/xhigh, small deterministic changes to Luna/medium, and mechanical read-only operations to Luna/low. - Third-Party Provider Registration: Optionally registers Responses API providers as CLI profiles with secure credential storage (macOS Keychain, environment variables, or auth helpers) and validates each profile before routing tasks to it. - Depth-Adaptive Planning: Escalates from initial planning to a single deep-planning pass when evidence shows high-risk signals like security boundaries, data migrations, or concurrency concerns. - Use Case: A user asks to refactor a legacy module across multiple files. The Skill first presents the model routing plan for confirmation, runs a Sol planning agent, detects cross-subsystem impact, adds one Sol/xhigh deep-planning pass, then hands the final plan to a Terra/xhigh executor and reports verified changes. ## Quick Start Ask the agent to plan and implement your coding task using the plan-execute-orchestrator, then confirm the proposed four-quadrant model routing before execution begins.
