networkx
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data sources and generative model outputs via the SandBase API, creating a vulnerability surface where malicious instructions could be embedded in the retrieved content.\n
- Ingestion points: Data is ingested through
sandbase_runandsandbase_run_getas described inreferences/sandbase-api-map.mdand the 'SandBase boundary' section ofSKILL.md.\n - Boundary markers: While the instructions emphasize "authorization boundaries" and "explicit assumptions," they do not define specific prompt delimiters or instructions for the agent to ignore potentially malicious content within external data.\n
- Capability inventory: The skill allows the agent to generate and execute local Python code (NetworkX) and call external capabilities via vendor-provided tools (
sandbase_run).\n - Sanitization: The workflow includes 'Quality gates' for inspecting data shapes, types, and units, which provides logical validation but not formal security sanitization of the input content.\n- [DYNAMIC_EXECUTION]: The skill involves the generation and local execution of Python code to perform graph analysis and data transformations.\n
- Evidence: The 'Workflow' section in
SKILL.mdinstructs the agent to produce and verify implementations or analysis artifacts, which necessitates runtime code generation and execution using the NetworkX library.
Audit Metadata