arize_toolkit

Use when writing or editing a workbench plan (.workbench/*.md), designing task graphs for parallel agent execution, or preparing work for the wb CLI to dispatch

Retrieve and debug trace data from the Arize ML observability platform. Use when users want to list recent traces, look up a specific trace by trace ID, get all spans within a trace, analyze trace performance (latency, tokens, cost), or export trace data. Triggers on "list traces", "show traces", "look at traces", "get traces", "trace ID", "show me the spans", "see the spans", "dig into a trace", "trace detail", "trace performance", "what traces", "debug trace", "span lookup", "trace latency", "trace tokens", "trace cost", "export traces". Prefer this skill over arize-toolkit-cli when the request is specifically about traces or spans.

Manage Arize platform resources using the arize_toolkit CLI. Use when users want to list, create, update, or delete Arize resources (models, monitors, prompts, evaluators, custom metrics, dashboards, spaces, organizations, users, or data imports) via the command line, configure CLI profiles, or perform platform administration tasks. Triggers on "list models", "create monitor", "arize_toolkit", "CLI", "arize cli", or any request to manage Arize platform resources from the terminal. Always use the "arize_toolkit" command, never "ax".

Query and analyze data from the Arize platform using GraphQL, or help build/validate GraphQL queries and mutations. Use when users want to explore spaces, models, monitors, datasets, any Arize platform data, OR when they need help writing, formatting, or debugging GraphQL queries/mutations for the Arize API.

Complete workflow for adding new GraphQL queries or mutations to the Arize Toolkit, including models, types, tests, client methods, and documentation. Use when (1) adding a new GraphQL query or mutation to arize_toolkit, (2) implementing new API functionality that requires queries, (3) user requests to add Arize API capabilities, or (4) user invokes /new-query-workflow. Triggers on phrases like "add a query", "create a mutation", "implement API for", "add support for [feature]", or explicit workflow invocation.