The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or persistence mechanisms were detected.
[COMMAND_EXECUTION]: The skill utilizes the ax CLI for standard operations such as listing, creating, and exporting experiments and datasets. These commands are within the expected scope for managing ML observability workflows.
[EXTERNAL_DOWNLOADS]: The skill references official and well-known Python packages (arize-ax-cli, openai, anthropic, google-genai) and connects to official vendor domains (arize.com, flight.arize.com). These sources are recognized as trusted.
[REMOTE_CODE_EXECUTION]: The skill provides a Python script template (infer.py) for the agent to generate and execute. This script is used for performing model inference on dataset examples, which is a legitimate task for evaluation workflows. The template uses standard libraries and relies on user-provided environment variables for authentication.
[CREDENTIALS_UNSAFE]: The skill includes explicit and robust security instructions to prevent the unsafe handling of API keys. It prohibits reading .env files, asking users to paste keys in chat, or hardcoding secrets, instead recommending the use of environment variables and managed CLI profiles.

arize-experiment