Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Python scripts for local multi-agent orchestration using a shared blackboard file, UUID-based permission tokens, and token budget tracking. Agents delegate tasks to specialized sessions, gate access to resource types like DATABASE and PAYMENTS, and validate results before committing. No network calls are made and no external credentials are required.
Runs a full multi-agent coordination stack locally without requiring any external API keys, cloud services, or network access.
When to use it
- Coordinating three specialized agents on a quarterly financial analysis
- Gating database access with justification-based permission checks before querying
- Tracking handoff budgets to prevent token waste across agent sessions
- Detecting silent agent failures with heartbeat health checks
- Running supervisor review before returning multi-agent results to a user
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Analyze Q4 revenue trends and generate a strategic recommendation
- 1Initialize task budget with swarm_guard.py budget-init
- 2Intercept each handoff with swarm_guard.py intercept-handoff to verify budget before delegating
- 3Send subtasks to data_analyst, risk_assessor, and strategy_advisor via sessions_send
- 4Read results from the shared blackboard and validate each with swarm_guard.py validate-result
- 5Run supervisor-review and commit final output only if verdict is APPROVED
A synthesized executive summary combining metrics, risk flags, and recommendations committed to the blackboard as task:001:final