arize-experiment

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest, process, and export model outputs (untrusted data) from various providers (OpenAI, Anthropic, Gemini). While no specific malicious payload is present, the workflow involves handling large volumes of external text that could contain instructions targeting the agent or the analysis pipeline.
  • [COMMAND_EXECUTION]: The documentation provides numerous shell command examples using ax, jq, and python. These are documented as the primary interaction method for the tool and do not involve obfuscated or hidden commands.
  • [SAFE]: The skill correctly advises users to manage credentials via environment variables ($ARIZE_API_KEY) rather than hardcoding them or passing them as raw flags, which is a security best practice.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:49 AM
Security Audit — agent-trust-hub — arize-experiment