skills.rest
wahajahmed010
wahajahmed010COMMUNITY
Explore all skills by this author →
GitHub

subagent-orchestration

Orchestrate OpenClaw subagents with sandbox-aware delegation and timeout tuning.

Updated May 21, 2026
Search Tags:#delegation#orchestration#timeouts#subagent#tool permissions#multi-agent council#sandbox constraints
One-click install
npx skills add https://github.com/wahajahmed010/openclaw-skills --skill subagent-orchestration-wahajahmed010
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: subagent-orchestration
Source: https://github.com/wahajahmed010/openclaw-skills/tree/main/skills/subagent-orchestration
Command: npx skills add https://github.com/wahajahmed010/openclaw-skills --skill subagent-orchestration-wahajahmed010

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This skill prevents slow failures and timeouts caused by subagent sandbox limits, missing tools, missing context, and overloaded gateway spawns.

Core Features & Use Cases

  • Agent-type aware orchestration: Choose Worker, Researcher, or Council patterns based on whether you need web access, file/code ops, or multi-model review.
  • Sandbox-constraint guardrails: Avoid unsupported operations like inline Python (python3 -c), unauthorized web tools, or reliance on the main session history.
  • Timeout and failure-mode playbook: Apply recommended runTimeoutSeconds, use lightContext: true, and map common symptoms (timeouts, gateway crashes, empty outputs) to concrete fixes.
  • Council synthesis pattern: Use a multi-model council approach by spawning three role-distinct subagents, then consolidating their verdicts (with a companion reference to council-of-llms).

Quick Start

Use subagent-orchestration to delegate your task by spawning a Researcher subagent with web tools allowed and a longer timeout, returning findings and sources.

Frequently Asked Questions about subagent-orchestration

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
Why do my delegated subagent tasks keep timing out or returning empty outputs?▼

Delegated subagent tasks time out or return empty outputs due to sandbox limits, missing tools, or overloaded gateway spawns. You can resolve this by applying appropriate runTimeoutSeconds, using lightContext, and ensuring correct toolsAllow permissions.

How do I orchestrate subagents for web-enabled research without hitting sandbox restrictions?▼

To orchestrate subagents for web research safely, spawn a Researcher subagent with web tools explicitly allowed via toolsAllow, use a longer runTimeoutSeconds, and enable lightContext to avoid unauthorized web tool errors and gateway crashes.

What is a multi-model council pattern for subagent delegation?▼

A multi-model council pattern spawns three role-distinct subagents to analyze a task independently, then consolidates their verdicts into a single synthesized output. This approach leverages diverse model perspectives for more comprehensive review.

Can I run inline Python scripts inside a spawned subagent sandbox?▼

No, you cannot run inline Python scripts like python3 -c inside a spawned subagent sandbox. The sandbox constraints block unsupported inline operations, so you must adjust your delegation rules to avoid them and prevent execution failures.

How do I choose between Worker, Researcher, and Council subagent patterns?▼

Choose your subagent pattern based on the task: Worker for file and code operations, Researcher for web-enabled research requiring web tools, and Council for multi-model review and analysis across different role-distinct agents.

What causes gateway crashes during subagent spawning and how can I prevent them?▼

Gateway crashes during subagent spawning are caused by overloaded spawns, missing context, and reliance on main session history. Prevent these crashes by enabling lightContext, tuning runTimeoutSeconds, and providing self-contained delegation rules.

Related Skills

Explore More Software Engineering Skills→
raphaelchristiraphaelchristi
43

harness:setup

Configure Harness Evolver, LangSmith, and baseline evaluations for Python projects.

Community
Advanced
boshu2boshu2
418

pre-mortem

Validate plans or specs with a multi-model council and structured verdicts.

Community
Advanced
47network47network
5

llm-council

Coordinate multiple LLMs through anonymous peer review and synthesis.

Community
Advanced
tzachbontzachbon
432

delegation-principle

Delegate Ralph workflow tasks to specialized subagents via the Task tool.

Community
Basic