Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Enables an agent to spawn specialized sub-agents that inherit its preferences and memory state. Supports parallel task decomposition via a swarm protocol, structured inter-agent signaling for clean data handoff, and git-based rollback to revert agent state to any prior cycle.
Structured swarm coordination with mandatory token-frugal constraints and git-based rollback avoids context overflow and irreversibility that single-agent approaches hit on large tasks.
When to use it
- Analyzing a large codebase across multiple files in parallel
- Distributing multi-file refactoring work across worker agents
- Rolling back agent memory and files after a failed experiment
- Injecting structured error context into a sub-agent for targeted repair
- Synthesizing parallel worker outputs into a single consolidated report
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Clone yourself to analyze the entire src/ directory
- 1Decomposes the src/ directory into 3-5 isolated sub-tasks by module or file group
- 2Spawns worker agents via sessions_spawn, one per sub-task
- 3Injects the Frugal Reading Protocol constraint into each worker's system prompt
- 4Workers complete analysis and fire structured SIGNAL: COMPLETE messages with payloads
- 5Master agent collects all signals and synthesizes a final report
A unified codebase analysis report compiled from parallel worker findings