Multi-Agent Research Coordinator Skill

Inspired by: Build a Multi-Agent Research Assistant

Purpose: Decompose research questions into parallel sub-tasks, dispatch them to specialized agents, and synthesize results.

What's Unique

A coordinator builds a dependency DAG of research tasks (web search, doc reading, summarization, critique, synthesis) and runs them in topological order. Each role runs against a least-loaded specialist agent.

Module: agents/research_coordinator.py

from agents.research_coordinator import (
    ResearchOrchestrator, ResearchAgent, AgentRole,
)

def web_handler(task, ctx): return "found X, Y, Z"

orch = ResearchOrchestrator()
orch.register_agent(ResearchAgent("web1", AgentRole.WEB_SEARCHER, web_handler))
# ... register agents for all roles ...
orch.plan("How does feature X work?")
orch.run()
print(orch.synthesize())

Default Plan Shape

web_search ──┐
             ├─→ summarize ─→ critique ─→ synthesize
doc_read  ───┘                  │
                                └────────────┘

Roles

• WEB_SEARCHER — external knowledge
• DOC_READER — internal docs / repo
• SUMMARIZER — distills findings
• CRITIC — gap analysis
• SYNTHESIZER — final answer

Quick-Note Issue: #238