The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon instructions found within the local repository environment (e.g., repository instructions, existing code conventions). While this represents a theoretical attack surface for indirect prompt injection, the skill includes significant guardrails.
Ingestion points: SKILL.md (Step 2: "repository instructions and conventions", "existing architecture and representative features").
Boundary markers: The skill contains a dedicated "Boundaries" section and mandates following existing project patterns rather than arbitrary directives.
Capability inventory: The agent is authorized to generate files (Step 5), modify configuration (Step 6), and execute build/test commands (Step 7).
Sanitization: The skill mitigates risks by instructing the agent to report decisions and alternatives rather than silently introducing third-party dependencies or refactoring unrelated code.
[COMMAND_EXECUTION]: The workflow incorporates standard .NET CLI operations such as dotnet restore, dotnet build, and dotnet test for code verification. These are legitimate, proportional, and necessary actions for a .NET development agent.

dotnet-scaffold