baoyu-image-gen
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documentation instructs the agent to use shell commands (
test -f) to verify configuration paths and utilizesnpxto execute the image generation script. - [EXTERNAL_DOWNLOADS]: The skill connects to official API endpoints for Google Gemini (
generativelanguage.googleapis.com) and OpenAI (api.openai.com) to perform its primary function. These are recognized as well-known and trusted services. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external files or standard input to use as prompts for image generation, which represents a potential vector for malicious instructions to be passed to the underlying AI models.
- Ingestion points: Prompt content is ingested through the
--promptfilesparameter and standard input inscripts/main.ts. - Boundary markers: There are no markers or system instructions used to delimit the external content from the prompt structure.
- Capability inventory: The skill has permissions to read local files, write the output image to disk, and make outbound network requests.
- Sanitization: The skill does not perform any sanitization or filtering on the ingested prompt strings before sending them to the service providers.
Audit Metadata