The Agent Skills Directory

Automated Production Code Modification: The skill's primary purpose is to modify C# production source code to introduce 'seams' (such as interfaces or delegates). While the instructions emphasize minimal and behavior-preserving changes, automated modification of production logic is a sensitive operation that requires manual review of the generated diffs to ensure no unintended side effects are introduced.
Indirect Prompt Injection Surface: The skill ingests untrusted data in the form of C# source code, .csproj, and .sln files from the user's workspace. There is a potential consideration where malicious content within these files (e.g., specially crafted comments) could attempt to influence the agent's behavior during the refactoring process. The skill does not explicitly define boundary markers for this ingested content.
Command Execution for Build and Test: The workflow requires the agent to build the affected projects and execute tests to verify the changes. This involves the execution of shell commands (e.g., dotnet build, dotnet test), which is necessary for the skill's functionality but represents a capability that should be monitored for unexpected parameters.
Encapsulation Bypass via InternalsVisibleTo: The skill may suggest adding InternalsVisibleTo attributes to production projects to allow test assemblies to access internal members. While this is a standard .NET testing pattern to avoid making testing hooks public, it bypasses standard class-level encapsulation and should be applied following the principle of least privilege as described in the workflow.
Ambient Dependency Overrides: The skill utilizes AsyncLocal<T> to manage static API overrides in a thread-safe and async-safe manner. The instructions provide robust patterns to prevent process-global state contamination, which is a positive security and stability practice for concurrent test execution.

testability-obstacle