The Agent Skills Directory

Verified Local Computation: The skill implements forecasting logic through standalone Python scripts (rolling_mean.py and batch_days_of_cover.py). These scripts perform simple mathematical operations (averages and ratios) on local datasets, which is a secure way to handle numerical analysis without relying on uncontrolled prompt-based calculations.
Structured Data Access: The scripts are designed to read from specific, local CSV files located at /mnt/user/data/. This represents standard internal data processing, and there are no indications of network exfiltration or access to sensitive user configuration files (such as SSH keys or environment secrets).
Context-Efficient Delegation: The skill defines a clear boundary for when to use a subagent, ensuring that high-volume historical data is processed in a separate context. This is a functional design choice for maintaining performance and does not introduce security risks, as the subagent operates within the same restricted environment.
Input Validation: The Python scripts include basic type casting (e.g., converting command-line arguments to integers) and use standard library modules (csv, json), minimizing the risk of unexpected execution behavior during data processing.

forecasting