nano-banana-pro-prompts-recommend-skill

Pass

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads categorized prompt data in JSON format from the developer's GitHub repository (YouMind-OpenLab/nano-banana-pro-prompts-recommend-skill) during initial setup and periodic updates.
  • [COMMAND_EXECUTION]: The skill executes its own maintenance script (scripts/setup.js) to synchronize the prompt library. It also utilizes curl to download sample images from the vendor's CDN (cms-assets.youmind.com) for preview purposes.
  • [PROMPT_INJECTION]: The skill ingests prompt templates from external JSON files, which is a surface for indirect prompt injection.
  • Ingestion points: Prompt templates are read from local files in the references/ directory, which are synchronized with a remote GitHub repository.
  • Boundary markers: None identified; the agent is instructed to use the template content directly as the base for generation.
  • Capability inventory: The skill has access to network operations (curl), script execution (node), and media transmission (message tool).
  • Sanitization: The skill relies on the curated nature of the source repository and does not perform active sanitization of the template content.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 12, 2026, 05:19 AM