network-data-analysis

Pass

Audited by Gen Agent Trust Hub on Aug 6, 2026

Risk Level: SAFE
Full Analysis
  • [DATA_EXFILTRATION]: The skill implements a strict allowlist for file access, restricted to specific directories such as ~/.openclaw/nsm/runs and ~/.openclaw/analysis. It explicitly documents the blocking of sensitive paths like .ssh, .aws, and .env through DuckDB's enable_external_access=false and lock_configuration=true settings.
  • [COMMAND_EXECUTION]: The skill enables SQL query execution via the analysis_query tool. The documentation specifies that the execution engine is in-memory and irreversibly locked down to prevent operations like INSTALL, LOAD, or COPY TO, which mitigates the risk of arbitrary command execution or persistence.
  • [PROMPT_INJECTION]: The instructions do not contain any patterns attempting to override agent behavior or bypass safety filters. Instead, they provide clear, technical guidelines for correct tool usage and data interpretation.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 6, 2026, 09:47 AM
Security Audit — agent-trust-hub — network-data-analysis