The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the management of AI experiments through the ax CLI, involving operations such as listing experiments, creating runs, and exporting data. It also includes workflows for executing inference scripts locally.
[DYNAMIC_EXECUTION]: The ax experiments run command allows for the execution of a Python task function defined in a local file (task.py). This is used to process dataset examples through custom logic or model APIs.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing the arize-ax-cli and various LLM provider SDKs (such as openai, anthropic, and google-genai) from official package registries.
[INDIRECT_PROMPT_INJECTION]: As the skill processes dataset content that may contain untrusted data, it maintains a surface for indirect prompt injection.
Ingestion points: External data enters the workflow via ax datasets export and is passed to inference scripts.
Boundary markers: The skill uses structured JSON for data handling, although specific sanitization of the content within those structures is left to the task implementation.
Capability inventory: The skill possesses the capability to execute CLI commands and local Python scripts.
Sanitization: The skill does not provide automatic sanitization for the content of processed dataset examples.
[SAFE]: The skill demonstrates a high level of security awareness regarding secret management. It explicitly instructs the agent to never ask for API keys in chat, to use environment variables for credentials, and to avoid reading sensitive files like .env for anything other than non-secret identifiers.

arize-experiment