The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data via a diagnostic script, creating a theoretical ingestion surface.
Ingestion points: The script scripts/linear_algebra_numerics_diagnostics.py reads data from a user-provided CSV file.
Boundary markers: None present in the data ingestion step.
Capability inventory: The script's capabilities are limited to reading a CSV, performing summary statistics (mean, standard deviation, correlation), and writing a JSON output file. It contains no network operations, subprocess executions, or dynamic code evaluation.
Sanitization: The script uses pandas.to_numeric and pandas.to_datetime with errors="coerce", which effectively filters out non-conforming content that might attempt to inject instructions, ensuring the agent only receives numerical summaries.
[COMMAND_EXECUTION]: The workflow involves running a local Python script for data processing. The script scripts/linear_algebra_numerics_diagnostics.py is a standard utility that uses argparse for parameter handling and pandas for data analysis. It does not invoke any shell commands or external processes.

linear-algebra-numerics