The Agent Skills Directory

[SAFE]: The skill provides a framework for building Azure Functions agents using the azurefunctions-agents-runtime package. No security issues were detected.
[EXTERNAL_DOWNLOADS]: The skill references the azurefunctions-agents-runtime package from PyPI and its source code in the official Azure GitHub organization. These are official sources for the skill's primary functionality.
[COMMAND_EXECUTION]: The skill provides instructions for the user to run standard developer tools such as azd (Azure Developer CLI), func (Azure Functions Core Tools), and pip to manage their project environment.
[REMOTE_CODE_EXECUTION]: The skill documents how to enable a Python code execution sandbox (execute_python) using Azure Container Apps Dynamic Sessions. This provides a managed and secure environment for agents to perform computations.
[PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection where external event data is ingested into the agent's context.
Ingestion points: Event data from Azure Function triggers (HTTP, Queue, Blob, etc.) is passed to agents as JSON payloads in the prompt.
Boundary markers: Trigger data is wrapped in JSON code blocks to distinguish it from instructions.
Capability inventory: Agents can utilize the execute_python tool and various Azure API connectors like Office 365 or Teams.
Sanitization: No explicit sanitization or filtering of incoming event data is described in the instruction files.

azure-functions-agents