The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill documents several methods for generating and executing code at runtime. In references/code-generation-printing.md, the lambdify function is shown converting symbolic expressions into executable Python functions using backends like NumPy or SciPy. The autowrap and ufuncify utilities are documented as tools to compile expressions into C, C++, or Fortran code using Cython or f2py, which is then imported and executed. Additionally, parse_expr is used to dynamically parse strings into symbolic expressions, a process that internally relies on Python's eval() function.
[DYNAMIC_EXECUTION]: The documentation in references/code-generation-printing.md provides examples for using pickle.load() to deserialize SymPy objects from files. Deserializing data from untrusted sources using the pickle module is a known security risk that can lead to arbitrary code execution.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by demonstrating how to ingest and process untrusted mathematical data from external strings, LaTeX, or Mathematica formats.
Ingestion points: The parse_expr, parse_latex, and parse_mathematica functions documented in references/code-generation-printing.md act as ingestion points for external data.
Boundary markers: None are defined in the provided examples; expressions are processed directly.
Capability inventory: The skill possesses capabilities for file system operations (via pickle), runtime compilation (autowrap, ufuncify), and dynamic function generation (lambdify).
Sanitization: The skill includes a text warning that users should 'validate and sanitize to avoid code injection vulnerabilities' when parsing input, though no specific sanitization code is provided in the examples.

sympy