The Agent Skills Directory

[Data Handling and Outbound Transfers]: The skill processes project source code and configuration to perform migrations. It explicitly notes that sending local tool results to a remote AI model involves outbound data, providing transparency for the user regarding data movement.
[Secret Management Practices]: Instructions are included to protect sensitive information, such as ensuring local.settings.json remains git-ignored and avoiding the inclusion of secrets in configuration changes. It further recommends using redacted inputs when source processing is restricted.
[Command Execution and Tooling]: The workflow involves using standard development tools like the .NET CLI (dotnet) and Azure Functions Core Tools (func) to restore dependencies, build projects, and run tests. These operations are conducted within the scope of the migration task and utilize the user's existing environment.
[Integration with Trusted Sources]: All external documentation and technical guides are sourced from official Microsoft domains (e.g., learn.microsoft.com), which are recognized as reliable and safe resources for software development.
[Managed Agent Delegation]: The skill provides a mechanism for handing off language-level updates to specialized agents. This process includes verification steps to ensure the target agent is official and requires explicit user permission before execution.
[Indirect Prompt Injection Surface]: Since the skill processes external code and project files (such as .csproj and Program.cs), a potential surface for indirect prompt injection exists. The skill mitigates this by instructing the agent to treat external content as data rather than instructions and to verify permissions before reading content. It also suggests using redacted or synthetic data for verification.

azure-functions-update