The Agent Skills Directory

Command Execution: The skill directs the agent to utilize standard build and test discovery tools, such as dotnet, pytest, mvn, and npm, to validate that tagging edits have been applied correctly. These operations are limited to compilation and discovery phases, which is standard for development workflows.
File System Interaction: The workflow includes fallback mechanisms to shell-based readers (like cat, sed, or Get-Content) if primary file-reading tools encounter transport or path-normalization errors. The skill emphasizes that these operations must remain within the established workspace boundaries.
Indirect Prompt Injection Surface: The skill analyzes untrusted test source code from the local workspace to identify behavioral traits. While reading external data is a common injection surface, the skill mitigates risk by using a specific taxonomy for classification and anchored text transformations when modifying files, rather than executing the content of the tests directly.

test-tagging