Orchestrate Complex Queries with Specialized Agents

Query Orchestrator is an autonomous agent that analyzes incoming queries, decomposes them into constituent tasks, and coordinates the optimal combination of specialized agents to deliver comprehensive solutions efficiently. It runs a six-step process: query analysis, parsing core objectives, explicit and implicit requirements, scope, complexity, and domain expertise needed; task decomposition, breaking the query into discrete tasks mapped to required skills, dependencies, and execution order; agent selection, matching tasks to specialist agents by capability, planning redundancy for critical tasks, and optimizing for parallel execution; coordination strategy, designing handoffs, success criteria, quality checkpoints, and contingency routes; execution monitoring, tracking progress, identifying bottlenecks, and making real-time adjustments; and integration and delivery, synthesizing multi-agent outputs, resolving conflicts, and formatting the final deliverable. It uses Read, Glob, Grep, WebSearch, and Bash tools and runs on the Opus model.

Use Query Orchestrator when a query is complex and multi-faceted enough to benefit from being broken into sub-tasks handled by different specialist agents in parallel or in sequence, and you want an explicit orchestration plan, a task breakdown, agent assignments, execution order, and quality gates, plus a progress-tracking report rather than one agent attempting everything at once. When the original query lacks specificity, it is designed to proactively clarify requirements before proceeding rather than guessing. It is an orchestration and coordination layer, not a specialist executor itself - it plans agent assignments and synthesizes their outputs, so it needs actual specialist agents available to hand tasks to.

Input is an incoming query, potentially complex or multi-faceted, with any available context on constraints or constituent requirements. Output is an Orchestration Plan - a query analysis with primary objective, secondary requirements, a 1-5 complexity level, and relevant domains; a task breakdown mapping each task to an agent type, expected output, dependencies, priority, and duration; agent coordination naming the primary and supporting agents, execution order, and quality gates; and deliverables with expected format, success metrics, and timeline - plus a Progress Tracking section listing execution status per task and any real-time workflow adjustments, resource reallocation, or risk mitigation actions.

Runs with Read, Glob, and Grep for reading and searching local files, WebSearch for external research, and Bash for running commands, on the Opus model - Opus is used for the higher-complexity orchestration reasoning this role requires. No external agent-coordination platform or API integration is documented; coordination is expressed through the orchestration plan and progress tracking output formats above.

Teams and workflows that need a single entry point to decompose a complex request into discrete tasks, assign each to the right specialist agent, track execution, and resolve conflicting outputs into one deliverable, rather than manually deciding which agent handles which part of a multi-domain task.