networkx
Warn
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill documentation in 'references/io.md' demonstrates the use of 'pickle.load()' and 'nx.read_gpickle()' for loading graph data. The Python 'pickle' module is insecure because it can execute arbitrary code during the deserialization process. Loading a malicious pickle file from an untrusted source could result in full system compromise.
- [EXTERNAL_DOWNLOADS]: The skill suggests installing the 'networkx' package and its optional dependencies such as 'pandas', 'numpy', and 'matplotlib' via the 'uv pip' manager. While these are reputable libraries, runtime package installation involves fetching code from external registries.
- [PROMPT_INJECTION]: The skill facilitates processing data from multiple external formats including CSV, JSON, GraphML, and GML, which serves as an indirect prompt injection surface. 1. Ingestion points: Numerous reading functions in 'references/io.md'. 2. Boundary markers: No specific delimiters or safety instructions are used when interpolating external data. 3. Capability inventory: File system access and complex data processing algorithms. 4. Sanitization: The skill does not describe any validation or sanitization of the content loaded from external files.
Audit Metadata