Agent Orchestration

• Analyzed: May 12, 2026, 3:15 AM
• Findings: 1
• Version: 1.0.0
• Hash: ff86bfc96d250ecc84bef9abba8328486d239010f0f8196a6c09a99670bd2f71
• Source: palm
• External report: View on VirusTotal

Findings (1)

The OpenClaw AgentSkills skill bundle 'agent-orchestration' is designed to improve AI agent prompting and management. It provides templates and instructions for various agent types (builder, research, review) and includes mechanisms for tracking agent performance and learning. The skill instructs the agent to interact with the file system (e.g., creating `notes/areas/active-agents.md`, `LEARNINGS.md`, and writing output files to specified paths) and execute internal OpenClaw commands like `sessions_list`. Builder agent templates also imply the ability to run shell commands (e.g., `npm install`, `node server.js`) and perform web searches for research agents. These capabilities are all aligned with the stated purpose of orchestrating and managing agents that perform tasks like building code, conducting research, or reviewing content. There is no evidence of intentional harmful behavior, data exfiltration, unauthorized persistence, or malicious prompt injection against the agent; the instructions in `SKILL.md` focus on structured and effective prompting.