What problem does it solve? When multiple test files or subsystems fail for unrelated reasons, fixing them sequentially wastes time and mixing investigations causes confusion. This Skill provides a structured pattern for splitting independent failures into focused agent tasks that run concurrently without interfering with each other. ## Core Features & Use Cases - Domain Identification: Groups failures by subsystem so each agent owns exactly one test file or module with no shared state. - Agent Prompt Template: Supplies a ready-made prompt structure with scope, constraints, and expected output for each dispatched agent. - Structured Handoff: Enforces a mandatory yaml:reconciliation block (agent_verdict, files_modified, tickets_created, unresolved_blockers) so the orchestrator can mechanically verify results against git status. - Use Case: Three test files fail with different root causes (tool approval, batch completion, abort handling). Dispatch three agents in parallel, each fixing one file, then review summaries and run the full suite to integrate. ## Quick Start Ask the AI to dispatch one parallel agent per failing test file, each with a scoped fix task and a mandatory reconciliation summary.
