The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/test_check_eval_overlap.py file uses subprocess.run to call the leakage detection script during unit tests. The call is implemented securely using a list of arguments rather than a shell string.
[INDIRECT_PROMPT_INJECTION]: The scripts/check-eval-overlap.py tool ingests training and evaluation data for overlap analysis. This processing of external text is a potential indirect injection vector if the resulting reports are later interpreted by an agent without proper boundary markers.

ml-engineering