comfyui-imagegen
Pass
Audited by Gen Agent Trust Hub on Feb 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses Python scripts and the
sessions_spawntool to execute background tasks. The provided usage examples include hardcoded file paths (e.g.,C:\\Users\\hal\\.openclaw\\workspace) and a specific Telegram recipient ID (595846104). - [EXTERNAL_DOWNLOADS]: The Python script communicates with a local ComfyUI API on
localhost:8188to submit jobs and download generated image data. It also interfaces with Telegram to deliver generated media. - [PROMPT_INJECTION]: The skill processes natural language input from users to generate image prompts, creating a surface for indirect prompt injection.
- Ingestion points: Natural language user requests processed in
SKILL.md. - Boundary markers: The agent is instructed to use a specific JSON schema for prompts, but lacks instructions to disregard control tokens or embedded commands within that structure.
- Capability inventory:
scripts/generate.pyperforms network requests and file system writes;sessions_spawnexecutes shell commands. - Sanitization: No sanitization of user-provided prompt strings is performed before inclusion in the JSON object or command-line arguments.
Audit Metadata