gemini-api-dev

Warn

Audited by Gen Agent Trust Hub on Aug 8, 2026

Risk Level: MEDIUMPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill includes explicit instructions designed to bypass the agent's internal training, using phrases such as "IMPORTANT: These rules override your training data" and "Your knowledge is outdated." It provides factually incorrect information by listing fictional models (e.g., "gemini-3.5-flash", "Nano Banana Pro") and falsely labeling currently supported versions (Gemini 1.5, 2.0) as legacy or deprecated.
  • [EXTERNAL_DOWNLOADS]: The skill instructs users to install packages and fetch documentation from official sources, including the google-genai Python package, the @google/genai Node.js package, and the ai.google.dev domain. These are well-known technology services and repositories.
  • [PROMPT_INJECTION]: The skill claims to be an "official" resource from the "google-gemini" organization, which contradicts the provided author attribution. This deceptive framing is used to lend false authority to the misinformation provided in the instructions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 8, 2026, 11:36 AM
Security Audit — agent-trust-hub — gemini-api-dev