gemini-api-agent-platform

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The content includes features that enable execution of model-supplied code and dynamic launching of third‑party packages (npx), which present clear remote code execution and supply‑chain abuse vectors that could be weaponized for backdoors or data exfiltration.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required workflow includes using the URL context tool (URL context tool retrieves and injects fetched page text into the model prompt), which can pull outsider-authored free text from arbitrary URLs at runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The example runs "npx -y @philschmid/weather-mcp" (args=["-y", "@philschmid/weather-mcp"]) which will fetch and execute remote npm package code at runtime and is used as an MCP tool the agent will call, meaning remote code controls runtime behavior.