networkx
Warn
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [DYNAMIC_EXECUTION]: The skill documentation in references/io.md demonstrates the use of pickle.load() and nx.read_gpickle() for graph serialization. The Python pickle module is not secure against maliciously constructed data and can result in arbitrary code execution during the unpickling process.
- [INDIRECT_PROMPT_INJECTION]: The skill provides extensive methods for importing graph data from external files and databases in references/io.md. This ingestion of untrusted data creates a surface for indirect prompt injection where malicious content in node or edge attributes could attempt to influence the agent behavior. 1. Ingestion points: Multiple file-reading functions like nx.read_edgelist, nx.read_graphml, and pd.read_sql_query in references/io.md. 2. Boundary markers: None present for external data ingestion. 3. Capability inventory: Includes file system writes (nx.write_*) and complex data processing capabilities. 4. Sanitization: No evidence of sanitization or validation of imported graph attributes.
- [EXTERNAL_DOWNLOADS]: The skill mentions installing the standard networkx package and its optional dependencies using uv pip install in SKILL.md.
Audit Metadata