The Agent Skills Directory

Command Execution: The skill employs shell commands like sed, cat, and Get-Content as fallback mechanisms for reading and editing files when primary tools are unavailable. It also utilizes standard ecosystem-specific commands such as dotnet build, pytest --collect-only, and mvn test-compile for verifying that source code modifications are syntactically correct. These operations are intended for local workspace management and are restricted to the current project context.
Indirect Prompt Injection: Because the skill analyzes the content of user-provided test files to determine their traits, there is an inherent attack surface for indirect prompt injection. Maliciously crafted test files could theoretically contain instructions designed to influence the agent's classification logic, although the skill includes specific heuristics and verification steps to maintain consistency.

test-tagging