The Agent Skills Directory

Indirect Prompt Injection: The skill processes existing test code and production code, as well as a repository-level overlay file (.agents/skill-overlays/dotnet-test/writing-mstest-tests.md). These serve as ingestion points where untrusted data could potentially influence agent behavior. While the skill lacks explicit boundary markers for code blocks, it validates overlay frontmatter and prioritizes user instructions. The skill has capabilities for file reading and command execution (dotnet test), but no explicit sanitization for ingested content is mentioned.
Command Execution: The skill workflow includes a verification step that uses the dotnet test command to ensure corrections are effective. Users should treat this as a standard development activity that executes project code, which is a necessary capability for a testing-focused skill.

writing-mstest-tests