gemini-interactions-api

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • Use of Official SDKs and Tools: The skill guides users to employ official, version-controlled libraries such as google-genai and @google/genai. It also leverages the gcloud CLI for secure environment configuration and service activation.
  • Authentication Best Practices: By recommending Application Default Credentials (ADC) via gcloud auth application-default login, the skill avoids the security risks associated with hardcoded API keys or manual secret management.
  • Data Processing Considerations: As the skill facilitates processing user input through an AI model, it introduces a potential surface for indirect prompt injection.
  • Ingestion points: Untrusted user content is ingested via the input parameter in interactions.create calls throughout SKILL.md.
  • Boundary markers: No specific boundary markers or instructions to isolate user input from system instructions are present in the examples.
  • Capability inventory: The skill demonstrates the ability to interact with the Gemini API and execute local tools (e.g., get_stock_price in the function calling section).
  • Sanitization: The provided examples do not include explicit input sanitization or validation routines.
  • While these patterns are standard for this type of functionality, users should ensure appropriate guardrails are in place when processing external data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 05:07 PM
Security Audit — agent-trust-hub — gemini-interactions-api