The Agent Skills Directory

[DYNAMIC_EXECUTION]: The documentation in references/io.md and SKILL.md describes using pickle.load() and nx.read_gpickle() to save and load graph data. These functions are known to be unsafe because they can execute arbitrary code during the deserialization process if the input file has been tampered with or comes from an untrusted source.
[DATA_INGESTION_SURFACE]: The skill facilitates reading complex graph data from numerous external file formats (GraphML, GML, Pajek, LEDA, CSV, SQL, etc.). While these are standard library features, they represent an indirect prompt injection surface where maliciously crafted graph data could attempt to influence the agent's behavior during analysis.

networkx