The Agent Skills Directory

Command Execution: The skill instructs the agent to execute shell commands such as dotnet build and dotnet test to verify code changes. While these are standard development operations, they involve running shell processes with arguments derived from the local workspace.
Indirect Prompt Injection Surface: The skill processes C# source code files which are considered untrusted data. Because the agent reads these files and has the capability to execute shell commands, there is a potential surface for indirect prompt injection if a source file contains malicious instructions (e.g., in comments) designed to influence the agent's behavior.
Ingestion points: Processes .cs files within the local workspace.
Boundary markers: No specific instructions are provided to the agent to disregard instructions found within code comments during the migration process.
Capability inventory: Shell execution (dotnet, sed, cat), file system write access, and package management.
Sanitization: The skill performs mechanical string replacement without specific sanitization for natural language instructions embedded in code.
File System Interaction (Shell Fallback): The workflow allows for the use of shell utilities like sed, cat, or Get-Content as a fallback for file manipulation. To mitigate risk, the skill explicitly requires the agent to verify that canonical paths remain within the workspace boundary and suggests using anchored replacements to ensure integrity.
External Package Management: The skill may add NuGet packages to the project (e.g., testing fakes). This involves downloading and integrating external code, which is a standard part of the migration workflow but remains an external dependency point.

migrate-static-to-wrapper