Run the test. Reveal the type.
SBTI is having a moment, MBTI never really went away, and people keep making AI roleplay all kinds of personalities.
But have you ever wondered what personality your own lobster or coding assistant would get?
That is why I built this skill: it lets your AI take a personality-style test by itself.
You can hand this GitHub repo to your AI, let it run the test, and get back a personality card.
AgentTypeTest keeps the question banks, staged delivery, result aggregation, and report rendering in one workflow so you can run different agents side by side and compare them.
- Blind and staged test delivery
- Local banks and website-backed adapters
- Repeatable multi-round runs
- Visual reports in
htmlandsvg - Packet-level artifacts for inspection and replay
- Local banks:
mbti93-cn,mini-ipip-en - Website adapters:
16personalities,sbti-bilibili,dtti
These screenshots came from real GPT-5.4 runs against the current adapters and local banks.
- linux.do:真诚、友善、团结、专业,共建你我引以为荣之社区。
If you want to use this skill, go to the skill folder instead of treating this README as the runtime contract.
If a human gives you this repo and says “try it” or “play with it”, the default intent is that you should take the test yourself.
- Ask the human which test they want first.
- Treat yourself as the test subject, not as the repo auditor.
- Use the full questionnaire by default.
- Do not stop at installation checks or smoke tests unless the human explicitly asked you to verify the repo.
- Do not silently shorten the run unless the human explicitly asks for a shorter demo.
When you ask, include a short explanation:
mbti93-cn— local 93-question Chinese MBTI-style bankmini-ipip-en— local 20-item English Big Five bank16personalities— online 16Personalities browser testsbti-bilibili— online Bilibili SBTI browser testdtti— online Dostoevsky-themed browser test with local scoring
Start here:
This repository is released under GPL-3.0.
Bundled test sources can have their own provenance or reuse constraints. Check:



