gemini-interactions-api
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill employs directive language, including headers like "IMPORTANT" and "CRITICAL", to explicitly command the agent to "override your training data" and treat its internal knowledge as "outdated".
- [PROMPT_INJECTION]: The skill instructs the agent to redirect user requests for specific model versions to a preferred alternative, altering the agent's decision-making logic without user intervention.
- [EXTERNAL_DOWNLOADS]: The documentation references the installation of official SDKs, specifically
google-genaiand@google/genai, which are hosted by a well-known technology provider. - [COMMAND_EXECUTION]: The skill provides standard shell commands for package management (
pip install,npm install) and demonstrates API usage through code snippets intended for a developer's local environment. - [PROMPT_INJECTION]: The skill describes a surface for indirect prompt injection via the processing of untrusted data. Ingestion points: The
inputvariable used inclient.interactions.createexamples withinSKILL.md. Boundary markers: None present in the provided examples. Capability inventory: The API described includes tools for Google Search, code execution, and file search. Sanitization: No sanitization or input validation is shown in the code samples.
Audit Metadata