vastai-host
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the management of GPU hosting operations through the
vastaiCLI tool. It defines numerous subcommands for listing machines, setting prices, and managing maintenance windows. The skill explicitly prohibits the agent from performing runtime installations (pip install) or using direct API calls, which is a recommended security practice. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a potential injection surface.
- Ingestion points: The
vastai reports <id>command inSKILL.mdretrieves data submitted by external renters. - Boundary markers: Absent. There are no instructions provided to wrap the output of renter reports in delimiters or to ignore potential instructions embedded within them.
- Capability inventory: The agent has the ability to execute
vastaisubcommands via the Bash tool, including disruptive actions likeunlist machineordelete machine. - Sanitization: Absent. The skill does not instruct the agent to sanitize or filter the content of renter-submitted reports before processing.
Audit Metadata