The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the dotnet CLI to build and run scripts, mkdir for creating temporary workspaces, and chmod for setting file execution permissions.
[REMOTE_CODE_EXECUTION]: Facilitates the compilation and execution of arbitrary C# code provided by the user or generated by the agent, which is a significant risk if the code originates from an untrusted source.
[EXTERNAL_DOWNLOADS]: Supports fetching third-party dependencies from NuGet using the #:package directive or dotnet add package command without integrity checks or source verification.
[PROMPT_INJECTION]: Contains a surface for indirect prompt injection as it processes and executes code based on untrusted user intent.
Ingestion points: C# code blocks and intent descriptions provided by users in the prompt context.
Boundary markers: Absent; there are no clear delimiters or instructions to prevent the agent from obeying embedded malicious commands.
Capability inventory: The skill can execute local files, modify file system permissions, and perform network operations for package management.
Sanitization: Absent; the skill does not perform any validation, sanitization, or safety scanning of the provided C# code before execution.

csharp-scripts