Parallel Thinker Skill

Purpose: Enable any agent to quickly obtain multi-perspective analysis by invoking specialized agents in parallel.

When to Use

Invoke this tool when:

• User asks complex, multi-faceted questions requiring strategic, analytical, financial, technical, or research insights
• Single-agent response would be too narrow
• You need to speed up response by parallelizing expert queries

How to Use

Call the parallel_think tool with:

• query: the original user question
• agents (optional): list of agent IDs to consult (default: ["strategist", "data-analyst", "finance", "expert-coder", "researcher"])
• synthesizer_agent (optional): agent ID to synthesize results (default: "synthesizer")

Example:

{
  "query": "Should we invest in gold now?",
  "agents": ["strategist", "data-analyst", "finance"]
}

What Happens

1. The tool sends the query to each specified agent in parallel (non-blocking)
2. Waits for all responses (timeout 30s per agent)
3. Forwards the query and all responses to the synthesizer agent for consolidation
4. Returns the synthesized answer to you

Implementation Details

• Uses OpenClaw CLI openclaw agent --agent --message
• Concurrency: up to 5 parallel calls (configurable via maxConcurrent)
• Timeouts: 30s per agent, 60s for synthesizer
• If any agent fails or times out, results are omitted and synthesizer proceeds with available answers