seo-image-gen
Warn
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires running a local installation script (
./extensions/banana/install.sh) and references external GitHub repositories (AgriciDaniel/banana-claude,YCSE/nanobanana-mcp) which are not associated with the identified vendor 'flexksx' or recognized trusted organizations. The content of the installation script is not included in the skill for review.- [METADATA_POISONING]: There is a discrepancy between the metadata author field ('AgriciDaniel') and the identified skill vendor ('flexksx'), which can be deceptive regarding the provenance and safety of the skill's components.- [COMMAND_EXECUTION]: The skill executes local Python scripts (scripts/cost_tracker.py,scripts/presets.py) for session management and relies on system tools like ImageMagick (magick) and FFmpeg for asset processing.- [INDIRECT_PROMPT_INJECTION]: User-provided descriptions are directly interpolated into the prompt construction logic for image generation, creating a vulnerability where malicious input could influence model behavior. - Ingestion points: User input captured via command arguments in
SKILL.md. - Boundary markers: The instructions do not define delimiters or defensive prompts to protect against instruction injection in user data.
- Capability inventory: Generation of images via MCP tools and file system modifications via ImageMagick and FFmpeg.
- Sanitization: There are no documented steps for validating or escaping user-supplied text before interpolation.
Audit Metadata