Full Analysis of OpenClaw clawdtributor Skill

I. Name Etymology Breakdown

II. Core Positioning

III. Full Feature List

1. Automatic Task Decomposition (Fan-out)

• Automatically identify tasks that can run in parallel and tasks with sequential dependencies
• Assign matching capabilities, models and permissions to each subtask
• Output a standardized subtask queue

2. Multiple Distribution Scheduling Strategies (4 built-in load balancing modes)

1. Round-Robin: Distribute tasks evenly to idle child Agents of the same type for balanced load
2. Auction Matching by Capability: Agents bid based on their skills, models and resources; tasks are assigned to the most suitable instance
3. Least-Busy: Prioritize assigning tasks to Agents with the fewest ongoing tasks to avoid single-point overload
4. Static Batch Fan-out: Spawn a batch of independent child Agents at once to run tasks synchronously (ideal for web crawlers, batch research, bulk code review scenarios)

3. Child Agent Lifecycle Management

• Dynamically create isolated child Agents: each with independent context, separate tool permissions and isolated model sessions to prevent cross-instance crashes and interference
• Automatic retry for timeout/failed tasks, redistribution of dead tasks, and isolation of faulty Agents
• Enforce a maximum concurrent Agent limit to avoid resource exhaustion

4. Result Aggregation & Merging (Fan-in)

• Deduplication, conflict validation and filling of missing data
• Auto-format outputs into a unified standard structure
• Optional built-in review child Agent for secondary accuracy verification of results

5. Cross-Instance Distributed Distribution (Advanced Feature)

IV. Distinction from clawtributor (Easily Confused)

V. Typical Application Scenarios

1. Batch research: Spawn multiple child Agents to search different competitors/materials simultaneously, then aggregate a comparative report
2. Bulk code review: Split multiple source files and assign them to child Agents for parallel auditing, output a unified vulnerability list
3. Large-scale web crawling: Split target URLs across multiple Agents for scraping, then merge all collected datasets
4. Complex writing pipelines: Parallelize research, material sorting, draft writing and proofreading to speed up production
5. Distributed AI production clusters: Load-balance task scheduling across an Agent pool spanning multiple servers

VI. Underlying Dependent Native Tools

• sessions_spawn: Create isolated child Agent sessions
• memory_shared: Shared memory for multi-agent exchange of intermediate data
• agent_list: Query all available worker Agents in the cluster
• task_queue: Local asynchronous task queue

VII. End-to-End Workflow Overview

1. The main Agent receives a complex requirement and calls clawdtributor
2. The skill decomposes tasks into a DAG and selects a scheduling strategy
3. Spawn child Agents in batches and assign independent subtasks for parallel execution
4. Monitor the status of every subtask; redistribute timed-out or failed tasks
5. Once all subtasks complete, aggregate, validate and merge all outputs, then return the full result to the main Agent

VIII. Deployment & Activation

• File path: repo/.agents/skills/clawdtributor
• Activation: Add it to the allowlist in the main Agent config file AGENTS.md: allow: ["clawdtributor"]
• Invocation methods: Natural language prompts (e.g. "Split this task into parallel subtasks for multiple child Agents to run and summarize") or slash command /clawdtributor run