The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the ax CLI to perform dataset operations (list, get, create, append, delete, export) and jq for manipulating JSON data. These commands are necessary for the skill's primary function.
[CREDENTIALS_UNSAFE]: While the skill manages ARIZE_API_KEY, it follows security best practices by instructing the agent and user to use environment variables, never pass raw keys as CLI flags, and avoid logging or echoing secret values in the session.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing its required CLI tool (arize-ax-cli) and data processing libraries (pandas, pyarrow) through official package registries like PyPI using pip or uv.
[SAFE]: The persistence mechanisms described, such as adding environment variables to shell profile files (~/.bashrc or ~/.zshrc), are standard practices for configuring developer tools and are used appropriately for storing non-sensitive configuration like the workspace name.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data (CSV, JSON, Parquet) from user-provided files or stdout. This creates a surface where malicious instructions within the data could theoretically influence the agent, though this is inherent to the skill's intended purpose of dataset management.
Ingestion points: ax datasets export --stdout (reading into context), ax datasets create/append --file (loading local files).
Boundary markers: None mentioned for the dataset content itself.
Capability inventory: Shell command execution (ax, jq), environment variable manipulation, and file system read/write.
Sanitization: Not specified for the data payload.

arize-dataset