cli-anything-openrefine

Pass

Audited by Gen Agent Trust Hub on Aug 4, 2026

Risk Level: SAFE
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill requires a local installation of the agent harness using python -m pip install -e . from the skill's own sub-directory. This is a standard installation procedure for a local tool and does not involve remote code downloads or execution from untrusted sources.
  • [DATA_EXFILTRATION]: Network operations are restricted to communication with the local loopback address (127.0.0.1) on port 3333. No external network requests or data exfiltration patterns were observed.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data files (e.g., CSV/TSV) and applies transformation rules, which is an inherent part of its functionality.
  • Ingestion points: Import of messy data files (e.g., messy.csv) via the project import command in SKILL.md.
  • Boundary markers: No specific delimiters or safety warnings for the agent are included in the instructions.
  • Capability inventory: The skill can read/write local files (data export) and interact with a local network server.
  • Sanitization: The skill relies on OpenRefine's internal processing logic for data transformation.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 4, 2026, 02:36 AM
Security Audit — agent-trust-hub — cli-anything-openrefine