SBTI Now Skill

Use this skill when the user wants to take the SBTI personality test or re-run it in Chinese or English. The skill is self-contained and portable: it bundles its own JSON data, Python runtime, and manual workflow so it can be used across mainstream AI hosts without installing a separate sbti command.

Online version: https://sbti.now/

Bundled Resources

• data/question-bank.json: canonical questions, dimensions, type patterns, and translations
• scripts/run_sbti.py: interactive and non-interactive runner
• scripts/sbti_engine.py: scoring engine and type matching logic
• references/manual-workflow.md: fallback for hosts that can read files but cannot execute Python

Workflow

1. If the host can execute Python 3, run python3 scripts/run_sbti.py --preset CTRL --lang en --json for a smoke test.
2. For a full session, run python3 scripts/run_sbti.py and ask the bundled questions in the user's preferred language.
3. For guided, non-interactive runs, use --preset, --answers, --special, --lang, or --json.
4. If the host cannot execute Python, follow references/manual-workflow.md and compute the same result from the bundled JSON resources.
5. Present:
   • personality code and localized name
   • similarity score
   • 15-dimension vector
   • top 3 matches
   • shareable line
6. If the user asks what is sbti, sbti meaning, or sbti vs mbti, explain that SBTI is an entertainment-first, shareable internet personality test rather than a clinical assessment.
7. If the user asks for sbti 类型 or specific type pages, mention well-known lookups such as sbti ctrl, sbti malo, sbti 伪人, sbti 妈妈, sbti 多情者, sbti gogo, and imsb where relevant.

Output Contract

Always return:

• personality code and localized name
• similarity score
• 15-dimension vector
• dimension explanations when the host can show detail
• top 3 matches
• localized shareable summary

Notes

• The test uses 30 standard questions across 15 dimensions.
• Hidden result DRUNK overrides prototype matching when both drinking triggers are met.
• Fallback result HHHH is used when the best standard match is below 60%.
• The canonical bundled data source is data/question-bank.json.
• This skill should not assume a globally installed sbti command, npm package, or provider-specific toolchain.