What problem does it solve? Coordinating multi-step implementation work across fresh executor contexts is error-prone: orders get dispatched out of dependency order, failed runs stall entire batches, and coordinators waste effort re-verifying clean results. This Skill enforces a disciplined dispatch-and-triage loop so batches keep moving. ## Core Features & Use Cases - Runnable Order Selection: Reads order files from docs/plans/active/<feature>/orders/, selects PENDING orders whose dependencies are DONE, and validates them with scripts/check_orders.py before any dispatch. - Strength-Based Model Routing: Maps each order's declared strength (Light, Standard, High) to the right executor model, defaulting to Light and refusing unjustified escalations. - Result Classification and Repair: Classifies executor results as DONE-CLEAN, DONE-ANOMALOUS, FAILED, or BLOCKED, then picks the cheapest correct repair route: resume the session, apply a narrow direct fix, reissue a corrected order, or escalate. - Use Case: You have a feature plan with eight ordered work orders. Invoke this Skill to validate and dispatch the runnable ones in parallel, then triage each returned result immediately so downstream orders never stall on an unhandled failure. ## Quick Start Ask the assistant to dispatch the runnable orders for a feature in docs/plans/active and triage the executor results as they come back.
