The Agent Skills Directory

[SAFE]: The skill performs all numerical analysis using Python scripts (h_refinement.py, dt_refinement.py, gci_calculator.py, and richardson_extrapolation.py) that use only the standard library math module, eliminating risks associated with external packages.
[SAFE]: Extensive input validation is performed on all command-line arguments, including conversion to floats, finiteness checks using math.isfinite(), and array length verification to ensure data integrity and prevent processing errors.
[SAFE]: No dangerous functions such as eval(), exec(), or patterns involving dynamic code generation or runtime compilation were detected.
[SAFE]: The skill instructions and scripts do not include any network operations, sensitive file access (e.g., SSH keys, credentials), or obfuscation techniques.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted numerical data through CLI arguments in its analysis scripts. Ingestion points include the --spacings, --timesteps, and --values arguments. While no boundary markers are explicitly defined in the prompt instructions, the capability inventory is limited to mathematical calculations and outputting JSON results. Input sanitization is effectively handled by strict float conversion and numeric range validation, preventing the data from being interpreted as instructions.

convergence-study