Agent Orchestration
v1.0.0Master the art of spawning and managing sub-agents. Write prompts that actually work, track running agents, and learn from every outcome. Part of the Hal Stack 🦞
⭐ 1· 2.7k·18 current·19 all-time
by@clawdnw
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description match the content: SKILL.md and templates provide prompt architectures, spawn/track procedures, and templates for builder/research/review agents. There are no unexpected credentials, third-party API requirements, or unrelated binaries requested.
Instruction Scope
Instructions explicitly tell agents to spawn sub-agents, run local commands (e.g., sessions_list, npm install, node server.js, kill), write and move files (notes/areas/active-agents.md, LEARNINGS.md, target output paths), and 'actually test' builds. This is coherent for an orchestration/build skill but grants the agent the ability to modify the workspace and execute arbitrary local commands. One template variation mentions 'Competitive Intel' with 'Weakness exploitation opportunities'—this is a potentially sensitive/abusable guidance item (not inherently malicious, but worth flagging).
Install Mechanism
Instruction-only skill (no install spec, no code files executed by registry). Lowest install risk. However, the runtime instructions assume local tools (node, npm, sessions_list) without declaring them; that's common for instruction-only skills but means the agent may attempt to run commands that aren't present.
Credentials
The skill declares no environment variables, secrets, or config paths. Templates reference local paths and common dev tools but do not request credentials or other unrelated secrets.
Persistence & Privilege
always:false and no install hooks. The skill can be invoked autonomously (normal platform default) but it doesn't request persistent platform-level privileges or modify other skills' configs.
Assessment
This skill is coherent for orchestrating sub-agents and providing templates, but it instructs agents to run local commands and create/modify files. Before installing or enabling autonomous execution: 1) only allow execution in a sandbox or non-sensitive workspace; 2) confirm whether the helper command sessions_list or other referenced tools exist where the agent will run; 3) disable autonomous invocation if you don't want the agent to spawn/execute workflows without human approval; 4) review and remove or adapt the 'Competitive Intel' variation if you want to avoid guidance that could be used for unethical purposes; 5) monitor filesystem and network activity the first few runs and restrict network or credential access as needed.Like a lobster shell, security has layers — review code before you run it.
latestvk97a7wbq1sjdbn8ds4k1wmh1nd80b77h
License
MIT-0
Free to use, modify, and redistribute. No attribution required.