skills/microsoft/learn-microsoft-agent-framework-with-foundry-zavashop-supply-chain-workshop/agent-framework-azure-ai-csharp/Gen Agent Trust Hub
agent-framework-azure-ai-csharp
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Dynamic Execution Surface: The framework features a
SubprocessScriptRunnerdesigned to execute local scripts, including Python and PowerShell, enabling modular skill expansion. This capability is managed through optional script approval settings, ensuring that automated actions can be gated by human review. - Indirect Prompt Injection Considerations: The skill allows agents to ingest data from external origins such as Bing Web Search, vector stores, and MCP tools. This creates an interface for processing untrusted content. To mitigate potential risks, the framework supports structured output formats and provides mechanisms for users to inspect and approve tool calls before execution.
- Command and Tool Invocations: Agents built with this framework can interact with local function tools and hosted services (e.g., Code Interpreter). These integrations allow the agent to perform computational tasks and API operations, which are central to its purpose as a functional assistant.
- Secure Credential Management: The documentation emphasizes the use of
DefaultAzureCredentialandAzureCliCredential, explicitly advising against hardcoded secrets or connection strings. This promotes industry-standard security practices for identity and access management.
Audit Metadata