The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were identified in the skill. The instructions and scripts focus on statistical experimental design and randomization logic.
[EXTERNAL_DOWNLOADS]: The skill specifies installation of Python packages including numpy, pandas, and pydoe from official repositories. These are standard tools for scientific computing and DOE.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface through user-provided factor names and labels. * Ingestion points: User-supplied factor names, stratum labels, and arm names are processed by scripts in randomization.py and doe_designs.py. * Boundary markers: No explicit boundary markers or 'ignore' instructions are used for user data. * Capability inventory: All scripts are restricted to mathematical operations using numpy and pandas. There is no evidence of subprocess execution, file writing outside of user-initiated to_csv calls, or runtime network activity. * Sanitization: Input data undergoes basic type and integrity validation (e.g., checking for finite numeric levels and unique labels) within the helper functions.

experimental-design