What problem does it solve? Choosing the wrong subagent — or doing everything in the main thread — floods the conversation context with verbose output and wastes turns. This Skill provides a disciplined four-step workflow (inventory, match, choose topology, delegate) for routing tasks to the right Claude Code subagent so bulky work stays out of the main context. ## Core Features & Use Cases - Agent inventory and inspection: List, search, and inspect built-in and custom agents (tools, model, scope) via a stdlib-only Python helper script. - Routing heuristics: Match tasks to the most specific agent by description, with default routings for Explore, Plan, and general-purpose. - Delegation patterns: Guidance on single agents, parallel fan-out, sequential chains, forks, and when to stay in the main thread, plus how to write a complete delegation prompt. - Use Case: You need to research how authentication, the database layer, and the API module work before a refactor. The Skill directs you to fan out three Explore agents in parallel, each returning only a summary, keeping search noise out of your main context. ## Quick Start Ask the assistant to find the right subagent for a task, for example: use the agent-finder skill to pick and delegate a codebase exploration of the auth module to the appropriate subagent.
