The Agent Skills Directory

[SAFE]: The skill consists of standalone Python utilities (loop_designer.py, hiring_calibrator.py, etc.) and informational markdown reference files. Technical analysis shows that the scripts are limited to data processing and local text/JSON output.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves ingesting and analyzing structured external data (JSON interview results and role definitions).
Ingestion points: Files are ingested through the --input argument in the three main Python scripts (loop_designer.py, question_bank_generator.py, and hiring_calibrator.py).
Boundary markers: Although explicit boundary markers for the agent are not defined in the markdown, the Python scripts act as a processing layer that validates and standardizes data before it would be presented to the agent.
Capability inventory: A review of all .py files confirms the absence of high-risk capabilities such as eval(), exec(), subprocess, or file writing to sensitive directories. All scripts use standard library modules like statistics, json, and math for their calculations.
Sanitization: The hiring_calibrator.py tool implements a _validate_interview_record method that strictly checks for the presence of required keys and validates that all scores are numeric and within the expected 1-4 range, effectively sanitizing the input data against common schema confusion or injection attempts.

interview-system-designer