arize-experiment

Warn

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Executes system shell commands for environment discovery, package installation, and experiment lifecycle management via the ax CLI tool.
  • [COMMAND_EXECUTION]: Performs modifications to user shell configuration files like .zshrc and .bashrc and sets persistent environment variables for credential persistence.
  • [EXTERNAL_DOWNLOADS]: Installs the arize-ax-cli package from Python package registries. This is a vendor-provided resource from the skill author.
  • [CREDENTIALS_UNSAFE]: Collects sensitive Arize API keys and Space IDs, persisting them locally in ~/.arize/config.toml and system environment variables.
  • [PROMPT_INJECTION]: Ingests potentially untrusted data from CSV, JSON, and Parquet files for experiment creation. Ingestion points: Run data files provided via the --file flag. Boundary markers: Absent. Capability inventory: CLI command execution and data export. Sanitization: Absent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 12, 2026, 10:22 AM