azure-ai-agents-persistent-java
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- External Library Dependencies: The skill utilizes standard Microsoft Azure libraries (com.azure:azure-ai-agents-persistent, com.azure:azure-identity) for agent interaction and authentication. These are official packages hosted on Maven Central.
- Indirect Prompt Injection Consideration: The SDK allows the agent to process messages provided by users. These inputs are interpolated into the agent's context, creating a potential surface for indirect prompt injection where data could be crafted to influence agent behavior. (1) Ingestion points: Messages are ingested via client.createMessage and CreateThreadOptions in SKILL.md and references/examples.md. (2) Boundary markers: The examples do not explicitly demonstrate the use of boundary markers for user-supplied content. (3) Capability inventory: The SDK supports tools like CodeInterpreter, FileSearch, and function calling. (4) Sanitization: No explicit sanitization or filtering of user messages is demonstrated in the provided code snippets.
- Agent Tool Capabilities: The examples demonstrate how to enable tools like CodeInterpreter and FileSearch. These tools provide the agent with powerful capabilities such as executing Python code in a sandbox or searching uploaded documents. Users should ensure that the agent's instructions and tool configurations are appropriate for their specific use case.
Audit Metadata