run-openclaw-agents
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill manages agent orchestration and sub-agent spawning through the
sessions_spawntool, which is regulated by strict task-definition requirements. - [DATA_EXFILTRATION]: Facilitates external communication via the
messagetool (supporting Discord, Slack, and Telegram), mitigated by mandatory user confirmation procedures that include content previews and irreversibility warnings. - [COMMAND_EXECUTION]: Provides hardware access capabilities (screenshots, GPS, camera) through the
nodestool, requiring a high-risk confirmation protocol and specific justification for each use. - [SAFE]: Implements a detailed risk management framework that categorizes orchestration tools by potential impact and defines required human-in-the-loop checkpoints.
- [SAFE]: Includes explicit architectural warnings regarding the ACP (Advanced Control Protocol) runtime, noting it executes on the host system and is unavailable within sandboxed environments.
Audit Metadata