gemini-api-dev

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). At runtime the skill explicitly instructs using fetch_url to retrieve documentation (e.g., https://ai.google.dev/gemini-api/docs/llms.txt and its linked pages such as https://ai.google.dev/gemmini-api/docs/function-calling.md.txt) which would be read and trusted as authoritative instructions and thus can directly control agent prompts/behavior.