Agent system tester
Describe your AI agent. Get a way to measure how well it works and compare changes.
Agent system tester is a free AI skill for the job "Test and fix code". Add it to Claude, ChatGPT, Gemini or Codex as a skill, or paste it into any chat.
Free · No account · CC BY 4.0 licence · Made by sickn33 · Level: expert · Thinking: high
Made by sickn33. See the original on GitHub; we copied it as it is, with its licence.
What it does
A way to measure how well your AI agent works, so you can tell if a change made it better or worse. Agents answer differently each run and can reach a goal by several paths, so one pass or fail test is not enough. The skill builds a rubric with several parts, such as accuracy, completeness, source quality and effort spent. It judges the final result and allows different routes to get there. An AI judge gives volume, and a person checks the tricky cases. It also suggests a realistic token budget, since that drives much of the result, and sets pass marks you can use as a gate before release.
You give it
- A description of what your agent does
- Example tasks and good answers (optional)
- The changes you want to compare
- Your token or cost limits (optional)
You get
- A scoring rubric with weighted parts
- A set of test tasks
- A judging plan using AI and human review
- Pass marks to use as a release gate
Use it when
- You changed a prompt and cannot tell whether the agent got better.
- You want to catch regressions before you ship an update.
- You are comparing two set-ups of the same agent.
- You need a pass mark an agent must hit before release.
How to use it
- Add it to your AI once: download it and add the folder to Claude, ChatGPT or Codex as a skill.
- Ask for the job in your own words. Your AI opens the skill when it fits.