The Agent Skills Directory

[SAFE]: The skill provides legitimate environmental analysis capabilities using established statistical methodologies like Mann-Kendall, Sen's slope, and STL decomposition. The code matches the stated ecological purpose.
[COMMAND_EXECUTION]: The skill executes local R and Python scripts (scripts/*.R, scripts/*.py) to perform analysis. These scripts are well-documented, include robust logging, and are limited to mathematical processing and generating plots. No suspicious subprocess calls or shell injections were detected.
[EXTERNAL_DOWNLOADS]: The documentation and scripts reference several well-known scientific libraries (e.g., pymannkendall, bfast, statsmodels, ruptures). These are standard dependencies in the data science and remote sensing communities.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it processes external time series data (CSV/GeoTIFF).
Ingestion points: Data is loaded via pd.read_csv in the Python scripts and read.csv in the R scripts.
Boundary markers: There are no explicit boundary markers or instructions to ignore potential commands embedded within the input data file content.
Capability inventory: The skill can write to the local file system (CSV outputs and PNG plots) but has no network or administrative capabilities.
Sanitization: The scripts enforce data type conversion (e.g., to float and Date), which naturally mitigates the risk of text-based instructions influencing the mathematical execution logic.

environmental-time-series