design

Pass

Audited by Gen Agent Trust Hub on Jun 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill implements its core logic via Python scripts that manage searching design data, constructing AI generation requests, and rendering HTML presentations for clients.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of established libraries such as google-genai and pillow, and communicates with official Google Gemini API endpoints for its multimodal generation capabilities.
  • [PROMPT_INJECTION]: The skill ingests untrusted user data (e.g., brand names, industry contexts, and design prompts) and interpolates it into instructions sent to LLMs.
  • Ingestion points: User-provided command line arguments in generate.py and search.py scripts.
  • Boundary markers: None identified; inputs are concatenated or formatted directly into prompt templates.
  • Capability inventory: The skill has the ability to write files to the local system and execute shell commands as described in the documentation for social media image exports.
  • Sanitization: Inputs used in filenames are slugified to prevent path traversal, but text used in LLM prompts is not escaped or filtered.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 25, 2026, 09:44 AM
Security Audit — agent-trust-hub — design