Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies
npx skills add https://github.com/oabdelmaksoud/Openclaw-skills-Compilations --skill dispatching-parallel-agents复制此命令并粘贴到 Claude Code 中以安装该技能
Use when managing the Claude CLI Proxy (ban protector) — start, stop, restart, status, logs, backup, test, setup, or register the proxy in OpenClaw. Triggers on: 'proxy status', 'ban protector', 'start proxy', 'stop proxy', 'proxy health', 'antiban', 'cli proxy'.
Run stateful multi-agent graph workflows using LangGraph (linear, supervisor, parallel, conditional). Use when tasks need dynamic routing, branching logic, or a supervisor agent delegating work. All LLM calls use OpenClaw's existing provider config — no API keys needed. Choose over autogen-collab (debate) and crewai-collab (fixed pipelines) when you need conditional edges, a supervisor making routing decisions, or parallel fan-out with synthesis.
Run multi-agent AutoGen-style debates when tasks need consensus across specialists. Uses OpenClaw's existing provider configuration — no API keys or extra setup needed. Use when: architecture decisions, design reviews, root cause analysis, or complex research where multiple expert perspectives improve the answer. Triggered automatically by Cooper for high-complexity tasks or explicitly via [autogen] tag or "debate this" / "get consensus" instructions.
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
Run structured multi-agent workflows using CrewAI (sequential, hierarchical, consensus). Each agent has a defined role, goal, and backstory. Tasks have explicit expected outputs. Use when: structured pipelines (research → design → implement), manager-supervised workflows, or consensus-building with clear acceptance criteria. All LLM calls use OpenClaw's existing provider config — no API keys needed.
This skill should be used when the user asks to 'demonstrate skills', 'show skill format', 'create a skill template', or discusses skill development patterns. Provides a reference template for creating OpenClaw skills.
| name | dispatching-parallel-agents |
| description | Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies |
| metadata | {"openclaw":{"emoji":"⚡"}} |
When you have multiple unrelated failures (different test files, different subsystems, different bugs), investigating them sequentially wastes time. Each investigation is independent and can happen in parallel.
Core principle: Dispatch one agent per independent problem domain. Let them work concurrently.
Use when:
Don't use when:
Group failures by what's broken:
Each domain is independent - fixing tool approval doesn't affect abort tests.
Each agent gets:
Use openclaw agent spawn to dispatch agents concurrently:
# Dispatch agents for independent problems
openclaw agent spawn --agent forge --task "Fix agent-tool-abort.test.ts failures"
openclaw agent spawn --agent forge --task "Fix batch-completion-behavior.test.ts failures"
openclaw agent spawn --agent forge --task "Fix tool-approval-race-conditions.test.ts failures"
Or in OpenClaw's orchestration context, dispatch via subagent mechanism.
When agents return:
Good agent prompts are:
BAD: Too broad: "Fix all the tests" - agent gets lost GOOD: Specific: "Fix agent-tool-abort.test.ts" - focused scope
BAD: No context: "Fix the race condition" - agent doesn't know where GOOD: Context: Paste the error messages and test names
BAD: No constraints: Agent might refactor everything GOOD: Constraints: "Do NOT change production code" or "Fix tests only"
BAD: Vague output: "Fix it" - you don't know what changed GOOD: Specific: "Return summary of root cause and changes"
Related failures: Fixing one might fix others - investigate together first Need full context: Understanding requires seeing entire system Exploratory debugging: You don't know what's broken yet Shared state: Agents would interfere (editing same files, using same resources)
After agents return: