agent-framework-azure-ai-py

Warn

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill directs users to install agent-framework and agent-framework-azure-ai via pip install --pre, identifying them as the 'Microsoft Agent Framework Python SDK'. However, these packages do not follow official Microsoft naming conventions. Furthermore, the code examples use inconsistent import names (agentframework in step 1 versus agent_framework in the 'Imported: Core Workflow' section), which is a common indicator of supply chain risk or potential typosquatting.
  • [COMMAND_EXECUTION]: Code examples demonstrate the use of HostedCodeInterpreterTool, which enables the agent to execute code within a hosted environment. This capability is high-risk and should be carefully managed when the agent interacts with external or untrusted data.
  • [PROMPT_INJECTION]: The skill describes an architecture where agents process data from external sources (web search and APIs) while having access to code execution tools, creating a surface for indirect prompt injection.
  • Ingestion points: Results from HostedWebSearchTool and MCPStreamableHTTPTool are ingested into the agent's context during the workflow described in SKILL.md.
  • Boundary markers: No delimiters or instructions to ignore embedded commands are present in the provided templates.
  • Capability inventory: The agents have access to HostedCodeInterpreterTool for execution and HostedWebSearchTool for network operations.
  • Sanitization: The skill does not include input validation or sanitization procedures for data retrieved from external tools.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 7, 2026, 09:22 PM
Security Audit — agent-trust-hub — agent-framework-azure-ai-py