azure-ai-projects-py
Pass
Audited by Gen Agent Trust Hub on Apr 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- Dynamic Code Execution Capabilities: The skill documents the use of the
CodeInterpreterTool, which allows agents to execute Python code within a sandboxed environment. This is a primary feature of the Azure AI Projects SDK, enabling data analysis and visualization. Additionally, theCodeBasedEvaluatorDefinitionallows developers to define custom evaluation logic as Python code strings. These features represent a powerful execution surface that should be managed according to the principle of least privilege within the Azure Foundry ecosystem. - Indirect Prompt Injection Surface: The skill includes patterns for processing external data via the
datasetsandindexesmodules, as well as a batch evaluation script (scripts/run_batch_evaluation.py) that loads JSONL files. As these datasets may contain untrusted content that influences agent behavior during evaluation or RAG operations, developers should consider using boundary markers or sanitization when interpolating this data into prompts. - Authentication and Credential Management: The skill correctly promotes the use of
DefaultAzureCredentialfrom theazure-identitypackage, which is a security best practice for Azure. This approach avoids the need for hardcoded secrets by leveraging managed identities or environment-based authentication. Configuration values like service endpoints are handled through environment variables, further reducing the risk of credential exposure in source code. - Enterprise Tool Integration: The documentation covers a wide range of integration tools including
McpToolfor Model Context Protocol servers andOpenApiToolfor REST APIs. These tools facilitate connections to external services. Theapi-reference.mdalso notes the existence of shell-based tools (LocalShellToolParam), which provide low-level system access if configured by the user. - Official Resource Verification: All referenced packages (
azure-ai-projects,azure-identity) and repository links (Azure SDK for Python on GitHub) originate from verified, official sources. The documentation is consistent with standard enterprise SDK patterns.
Audit Metadata