Better LLM prompts, grounded in real failures.
Improves LLM prompts by using production failures and evaluation feedback to produce revised prompts and measure their performance.
When to use it
Use when asked to optimize or debug prompts, improve AI responses, or improve output quality.
Give it a prompt-improvement request; it uses Arize performance data to return revised prompt messages and explain the changes.
Requires the ax CLI at version 0.14.0 or higher; troubleshooting provides installation and upgrade instructions.
Requires access to Arize through a configured profile and a valid API key for the relevant space.
This skill
ax profiles
Creates or updates Arize profiles (irreversible)
Arize
Creates Arize prompt experiments