Turn production failures into stronger prompts.
prompt engineering
Improves LLM prompts by analyzing production traces, evaluations, annotations, datasets, and experiment results.
When to use it
Use when the user wants to optimize, debug, tune, or improve an LLM or system prompt.
Ask it to optimize a prompt; it returns revised prompt messages and a brief explanation of the changes.
This skill
Arize profile
Updates your Arize profile (irreversible)
LLM provider
Sends prompt data to an LLM
Arize experiments
Starts Arize experiments
Arize
Reads your Arize performance data
Requires ax 0.33.0 or higher.
Requires an Arize profile configured for the CLI.
Uses jq to extract prompts, join records, find failures, and compare experiment scores.
The Arize API key can be loaded from a saved CLI profile.