networkx
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill documents the use of the
picklemodule for graph serialization and deserialization inreferences/io.md. Specifically, the example forpickle.load()shows a pattern that can execute arbitrary code if the input file is not from a trusted source. The skill mitigates this risk by providing a clear security warning to only unpickle files from trusted sources.\n- [INDIRECT_PROMPT_INJECTION]: The skill facilitates graph data ingestion from numerous external sources, including GraphML, JSON, CSV, and SQL databases, which creates an indirect prompt injection attack surface.\n - Ingestion points: Data is read via functions such as
read_graphml,node_link_graph,from_pandas_edgelist, andpd.read_sql_queryin thereferences/io.mdandSKILL.mdfiles.\n - Boundary markers: The skill does not explicitly provide instructions for using boundary markers or delimiters for the content of graph data objects.\n
- Capability inventory: The skill includes functions for file writing (
write_graphml,plt.savefig,pickle.dump) and database interaction (to_sql).\n - Sanitization: The skill recommends the use of parameterized SQL queries to prevent SQL injection vulnerabilities during database integration.
Audit Metadata