visual-creator
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill connects to the ByteDance Volcengine (Doubao) API at
ark.cn-beijing.volces.comto generate visual content. It subsequently downloads the generated images via HTTP GET requests for local processing and cropping. As this targets a well-known service for the skill's primary function, it is considered safe. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided strings for titles and subtitles which are then interpolated into image templates and filenames. While this creates a surface for indirect injection, the risk is minimal given the visual rendering context.
- Ingestion points: User-provided
titleandsubtitlevariables incover_generator.pyandhandler.py. - Boundary markers: The skill does not implement specific boundary markers or delimiters for the injected text within its template engine.
- Capability inventory: The skill has capabilities to write to the local
output/directory and specified projectassets/images/folders, and performs network requests to the image generation API. - Sanitization: The skill performs standard filename sanitization via the
sanitize_filenamefunction to ensure generated files are compatible with the local filesystem, though no specific sanitization is applied to the text rendered on the images.
Audit Metadata