nanobanana-infographic
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data (infographic briefs) to construct prompts for an external image generation model. This creates a surface where instructions embedded in the brief could influence the agent's behavior.
- Ingestion points: Data enters through the infographic brief structure (e.g.,
templates/brief.json). - Boundary markers: The skill provides explicit text rules (e.g., "Title max 5 words", "No other text") and composition constraints to limit model improvisation.
- Capability inventory: The skill includes scripts for file writing and network communication with the Google Gemini API.
- Sanitization: Input data is normalized (e.g., stripping whitespace) in
scripts/build_variant_pack.py. - [COMMAND_EXECUTION]: The skill utilizes local Python scripts (
scripts/probe_gemini_image_api.py,scripts/render_variant_pack.py) to automate the generation of prompts and the rendering of images. - [EXTERNAL_DOWNLOADS]: The skill makes network requests to the Google Gemini API (
generativelanguage.googleapis.com) to generate and fetch image data. As this targets a well-known and trusted service for the skill's primary purpose, it is considered a safe operation within the context of the skill.
Audit Metadata