The Agent Skills Directory

[SAFE]: The skill facilitates a governed workflow for knowledge graph benchmarking through deterministic Python scripts that handle initialization, validation, and evidence packaging.
[SAFE]: All scripts use secure coding practices, such as employing yaml.safe_load for parsing external artifacts and utilizing SHA-256 hashing to ensure data integrity across the specification and run packages.
[SAFE]: The system includes a proactive security feature (scan_boundary_leakage in mts_common.py) that audits processed data for prohibited keys and phrases, effectively preventing prompt injection or schema confusion that could introduce subjective quality judgments into the measurement layer.
[SAFE]: The skill correctly segregates design-time intent from runtime execution facts and evidence, maintaining clear data lineage and preventing persistence or state manipulation attacks.
[SAFE]: No unauthorized network connections, access to sensitive local environment paths (e.g., credentials), or obfuscated commands were identified in any of the provided files.

fin-kg-define-measurement-test