The Agent Skills Directory

[COMMAND_EXECUTION]: The documentation in references/io.md illustrates the use of pickle.load() and nx.read_gpickle() for saving and loading graph objects. Loading pickled data is a known security risk that can allow an attacker to execute arbitrary Python code by providing a specially crafted file.
[PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data, which presents a surface for indirect prompt injection attacks.
Ingestion points: references/io.md describes methods for reading graph data from various formats including GraphML, GML, JSON, CSV, and SQL databases.
Boundary markers: No delimiters or instructions to ignore embedded commands are present in the documentation.
Capability inventory: The skill uses networkx, pandas, and numpy for data manipulation and analysis.
Sanitization: There is no evidence of input validation or content filtering for the data being read.
[EXTERNAL_DOWNLOADS]: SKILL.md contains instructions for installing the networkx package and its dependencies using pip. This is standard practice for software development tools.

networkx