swarm-workflow-protocol

What This Skill Does

The swarm-workflow-protocol is the foundational operating system for multi-agent collaboration within the 0x-wzw swarm. It defines the rigid yet adaptive architecture for how agents spawn, communicate, and hand off tasks. Unlike traditional hierarchical automation, this protocol promotes a 'humans spar, agents drive' philosophy, removing approval bottlenecks to achieve full autonomy through continuous improvement. The protocol mandates a strict Pre-Task Spawn Analysis to ensure computational efficiency, preventing unnecessary token overhead. It includes a robust relay communication system using dedicated endpoints for message passing between agents, ensuring seamless coordination across complex, parallelizable tasks.

Installation

To integrate this protocol into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/0x-wzw/swarm-workflow-protocol

Ensure that your local relay server (defaulting to localhost:18790) is active and configured with the required x-auth-token: agent-relay-secret-2026 to allow for inter-agent message propagation.

Use Cases

• Large-Scale Data Synthesis: When an ultra-complex task requires decomposing data across multiple independent analytical seams.
• Autonomous Research Pipelines: Enabling agents to research, debate, and synthesize findings without waiting for explicit human confirmation for every micro-decision.
• Complex Development Workflows: Coordinating multiple specialist agents to handle distinct modules of a codebase simultaneously.

Example Prompts

1. "Perform a pre-spawn analysis for the Q4 audit report. Evaluate the complexity and verify if we have sufficient parallel seams to justify a multi-agent approach."
2. "I am implementing the new data scraper logic. Here is my reasoning for the architecture: [details]. Instead of asking for approval, challenge any technical gaps you see in my approach."
3. "Initiate a task delegation to the research agent; transmit the preliminary findings and wait for the synthesis report via the relay."

Tips & Limitations

• Token Economy: Always perform the 'Token Math' check. If the overhead of spawning a new agent exceeds the value of the output (a 3-5x return on investment), default to serial execution within the main session.
• Sparring Culture: Shift your mindset from 'approver' to 'sparring partner.' Your goal is to sharpen agent decisions, not serve as a gatekeeper.
• Resistance: Even when a task seems complex, resist the 'swarm urge' to spawn agents for tasks with many serial dependencies, as this creates unnecessary latency.