This repository turns the terminal-style SBTI flow into a portable skill root: bundled question bank, local Python scoring, multilingual prompts, and a manual fallback when the host cannot execute code.
It is built for hosts and agent runtimes that want a drop-in SBTI skill without fetching extra data or depending on the sbti npm CLI.
The repository already includes SKILL.md, index.json, agents config, and bundled data, scripts, and references so skill installers can consume it directly.
Question data lives in data/question-bank.json, while run_sbti.py and sbti_engine.py keep execution and scoring local and predictable.
The skill covers Chinese and English prompts and supports lookups such as sbti ctrl, sbti malo, sbti 妈妈, sbti 多情者, and imsb.
If the host can read files but cannot run Python, references/manual-workflow.md preserves the same test flow without extra tooling.
These are the questions that come up when a host wants to add an SBTI skill instead of rebuilding the test flow from scratch.
Use the GitHub repository if you want the portable skill package, or run the local test flow here to see the same SBTI experience in action.