The Agent Skills Directory

[SAFE]: The skill operates as a local, network-free auditor for model provenance records. The provided Python script (scripts/check_model_provenance.py) uses only standard libraries and performs simple set arithmetic and token matching to identify potential conflicts between model training and evaluation data. It does not ingest untrusted remote data, spawn subprocesses beyond its own test suite, or access sensitive system files. All tests and challenge scripts use synthetic, PII-free data fixtures.

model-sourcing