controlnet-pose
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the RunComfy command-line interface from the official NPM registry, a well-known package service.
- [COMMAND_EXECUTION]: Executes the
runcomfyCLI tool. Access is appropriately restricted to this specific tool through the skill's configuration. - [INDIRECT_PROMPT_INJECTION]: The skill processes external media URLs to guide AI generation, which presents a theoretical attack surface for indirect prompt injection.
- Ingestion points: Video and image URLs are accepted through the
--inputJSON argument inSKILL.md. - Boundary markers: The documentation includes instructions for agents to only process URLs explicitly provided by the user and provides guidance on identifying divergent output.
- Capability inventory: The
runcomfytool performs network operations to reach vendor APIs and writes generated media to the local filesystem. - Sanitization: The tool uses JSON strings for input to prevent shell command injection and limits outbound traffic to the vendor's domains.
Audit Metadata