The Agent Skills Directory

[SAFE]: The skill is focused on spatial conservation planning and utilizes well-known, established scientific R packages such as prioritizr, terra, and sf. No malicious instructions, obfuscation, or persistence mechanisms were detected.
[COMMAND_EXECUTION]: The skill invokes local R scripts (run_prioritization.R and prioritization_sensitivity.R) to perform its analytical tasks. These scripts are executed via the Rscript command and utilize standard library functions to process data; they do not perform unsafe shell operations or execute arbitrary strings.
[DATA_EXFILTRATION]: All data processing, including reading input rasters/CSVs and writing results, occurs within the local file system. The analysis of the R scripts and skill instructions confirms a lack of network-based exfiltration or unauthorized access to sensitive files like credentials or environment variables.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data files (GeoTIFF rasters and CSV tables). While these represent ingestion points for untrusted data, the risk is negligible due to the structured, binary (raster), and numeric (CSV targets) nature of the inputs, which are used for mathematical optimization rather than natural language processing. The scripts perform basic existence and spatial consistency checks on these inputs.

spatial-prioritization