The Agent Skills Directory

[SAFE]: The skill provides scientific calculation tools that incorporate strong security controls and defensive programming practices.
[REMOTE_CODE_EXECUTION]: To prevent arbitrary code execution, the skill implements a custom expression parser in _common.py. This parser converts mathematical strings into an AST and validates every node against a strict whitelist of operations and functions. This approach ensures that user-supplied measurement models cannot execute malicious system-level code.
[COMMAND_EXECUTION]: The skill uses bundled Python scripts executed via the Bash tool. These scripts are designed for local operation only, incorporate input size limits (e.g., MAX_INPUT_BYTES), and use atomic, private-permission file writes to prevent race conditions and data exposure.
[DATA_EXFILTRATION]: The skill implements a checked_input_file helper that explicitly validates file paths to reject network URLs and symbolic links. This mitigates risks associated with local file inclusion and ensures the agent cannot be tricked into reading sensitive system files outside the intended scope.
[EXTERNAL_DOWNLOADS]: The skill references well-known, authoritative scientific libraries including pint, uncertainties, numpy, and scipy. These dependencies are consistent with the skill's primary function and are installed from standard package registries.

uncertainty-and-units