The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute various local shell scripts and commands such as ./repo.sh, make, and python3 -m unittest. These commands are standard for building and testing software within a local repository context and do not involve fetching or executing code from external or untrusted sources.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing data from the local environment, specifically file paths from git diff and logs from CI failures. This constitutes an ingestion point for external data that is then interpolated into shell commands. However, this surface is limited to the local development context and represents the intended core functionality of a developer-oriented test selection tool. There are no explicit boundary markers or sanitization steps mentioned, but the risk is assessed as safe given the specific use case and execution environment.
[DYNAMIC_EXECUTION]: The instructions involve setting environment variables like LD_LIBRARY_PATH and PYTHONPATH to point to local build artifacts (_build/linux-x86_64/release). This is a standard and necessary practice for linking shared libraries and discovering modules during the testing phase of C++/Python integrated projects.

usd-brep-test-selection