supervisor-api
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill uses official, versioned packages from the vendor (databricks-openai and databricks-sdk) for core functionality.
- [SAFE]: Provides instructions for managing configuration and secrets using environment variables in .env files, which is a security best practice.
- [INDIRECT_PROMPT_INJECTION]: The skill implements a handler that processes user messages to perform tool orchestration. This represents an inherent attack surface for indirect prompt injection where external data enters the LLM context. However, since the skill utilizes official platform SDKs and the behavior is fundamental to the documented API's purpose, the finding is categorized as safe.
- Ingestion points: agent_server/agent.py via the request.input field in invoke_handler and stream_handler functions.
- Boundary markers: None are explicitly defined in the boilerplate code to delimit user input from system instructions.
- Capability inventory: The agent is configured to interact with Databricks hosted tools including Unity Catalog functions, Genie spaces, and Knowledge Assistant endpoints, as well as client-side Python functions.
- Sanitization: No explicit sanitization or filtering of the request.input content is performed before passing it to the client.responses.create method.
Audit Metadata