← Back to Skills Marketplace
Subagent Orchestration
by
Wahaj Ahmed
· GitHub ↗
· v1.3.0
· MIT-0
137
Downloads
0
Stars
0
Active Installs
5
Versions
Install in OpenClaw
/install subagent-orchestration
Description
Orchestrate OpenClaw subagents effectively. Covers all three agent types (Worker, Researcher, Council) with correct spawning patterns, sandbox constraints, t...
Usage Guidance
This skill appears to do what it says (orchestrate subagents), but exercise caution before installing and using it:
- Do not paste secrets or private conversation history into the task parameter or inline context; prefer sanitized summaries.
- Treat any subagent that can exec and read/write files as high privilege; run orchestration experiments in an isolated account or ephemeral environment first.
- Only grant web/tool access (toolsAllow) to Researcher subagents when absolutely necessary; restrict which models or endpoints they can call.
- Verify the provenance of companion skills and referenced GitHub repos (README points to 'wahajahmed010' while the registry owner differs). Clone and inspect those repos before installing.
- Review and limit config defaults (runTimeoutSeconds, maxConcurrent) to reduce blast radius, and prefer lightContext:true as recommended.
If you want, I can extract the exact lines that instruct pasting full context or summarize what would be exposed by following the guide step-by-step.
Capability Analysis
Type: OpenClaw Skill
Name: subagent-orchestration
Version: 1.3.0
The subagent-orchestration skill provides documentation and templates for managing OpenClaw subagents, focusing on Researcher, Worker, and Council patterns. It includes guidance on using the sessions_spawn tool, managing timeouts, and configuring tool permissions (toolsAllow). The content is strictly educational and operational, aimed at improving agent reliability and avoiding common failure modes like context overflow or gateway timeouts, with no evidence of malicious intent, data exfiltration, or unauthorized execution.
Capability Assessment
Purpose & Capability
Name/description (subagent orchestration) align with the runtime instructions: sessions_spawn patterns, sandbox constraints, timeout rules, and use of web-enabled 'Researcher' agents. Recommended models and companion 'Council of LLMs' use are consistent with a multi-model council pattern. Minor provenance inconsistency: README/SKILL.md recommend installing 'wahajahmed010/...' companion skills while the registry Owner ID for this package is different; verify upstream source.
Instruction Scope
SKILL.md explicitly tells the agent to spawn subagents that can read/write files, execute scripts, and (when allowed) access web tools. It also instructs the agent to 'paste context inline' because council agents don't have the main conversation history. That guidance is coherent for delegation, but it increases risk of sensitive-data exposure: pasting session history or secrets into tasks or files handed to subagents could leak information to less-trusted subagents or any model with web access. The instructions also encourage creating and handing .py scripts to subagents and running shell commands — valid for orchestration but a high-privilege action if subagents are untrusted.
Install Mechanism
Instruction-only skill with no install spec and no code files — nothing is written to disk by the skill itself. This lowers installation risk. The README suggests installing companion skills from external sources (GitHub/ClawHub) — the user should verify those external repositories before installing.
Credentials
The skill requests no environment variables, no binaries, and no config paths beyond recommending user-editable OpenClaw config files (~/.openclaw/openclaw.json and ~/.openclaw/council-config.json). Those config edits are reasonable for a subagent orchestration skill. However, the skill's allowed operations (file read/write, exec) mean that any secrets present in the working directory or in files could be accessible to subagents, so the lack of declared secrets does not eliminate risk of accidental exposure.
Persistence & Privilege
The skill does not request 'always: true' or any special persistent privileges. It instructs editing user config files under the user's home (reasonable for defaults). It does not modify other skills or system-wide settings beyond user-level config.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install subagent-orchestration - After installation, invoke the skill by name or use
/subagent-orchestration - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.3.0
Enhanced description, added tags, deepseek-v4-pro model, companion skill links
v1.2.1
Update README with multi-model council pattern, companion skill links, deepseek-v4-pro
v1.2.0
Add multi-model council pattern, deprecate single-model councils
v1.1.0
Added timeout config defaults, gateway timeout troubleshooting, runTimeoutSeconds to spawn patterns
v1.0.0
Initial release
Metadata
Frequently Asked Questions
What is Subagent Orchestration?
Orchestrate OpenClaw subagents effectively. Covers all three agent types (Worker, Researcher, Council) with correct spawning patterns, sandbox constraints, t... It is an AI Agent Skill for Claude Code / OpenClaw, with 137 downloads so far.
How do I install Subagent Orchestration?
Run "/install subagent-orchestration" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Subagent Orchestration free?
Yes, Subagent Orchestration is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Subagent Orchestration support?
Subagent Orchestration is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Subagent Orchestration?
It is built and maintained by Wahaj Ahmed (@wahajahmed010); the current version is v1.3.0.
More Skills