What problem does it solve? Investigating multiple unrelated failures sequentially wastes time and bloats your context. This Skill provides a structured pattern for delegating independent problems to isolated agents that work concurrently, each with focused scope and self-contained instructions. ## Core Features & Use Cases - Parallel Agent Dispatch: Launch one agent per independent problem domain (e.g., one per failing test file) so investigations run concurrently instead of sequentially. - Agent Type Selection: Match work to the right agent — read-only agents like codebase-locator and codebase-analyzer for investigation, general-purpose agents for read-write fixes. - Prompt Structure Guidance: Craft focused, self-contained agent prompts with clear scope, constraints, and expected output to prevent agents from drifting or over-editing. - Use Case: After a refactoring leaves 6 test failures across 3 files, dispatch 3 agents in parallel — one per file — then review summaries, verify no conflicts, and run the full suite to integrate all fixes. ## Quick Start Ask the assistant to dispatch one parallel agent per failing test file to investigate and fix each failure independently, then review and integrate the results.
