SBTI 人格测评

SBTI（沙雕大型性格指标）人格测评工具。由B站UP主@蛆肉儿串儿创作，31道题，15维度，27种人格类型。 当用户提到：(1) SBTI测试、(2) 沙雕人格测试、(3) 帮我做SBTI测评、(4) 测测我是什么人格、(5) SBTI答题、(6) 指定人格类型答题。 不在用户仅提到"MBTI"、"人格测试"等模...

Security Scan

The skill is mostly coherent for delivering a personality quiz, but there are mismatches between the documentation and the included script (hidden Q31/Q32 and DRUNK handling) and a small undeclared dependency (python3), so review before use.

Purpose & Capability

Name, description, and files (questions, types, calculate.py) align with a personality-quiz skill. However SKILL.md shows examples invoking `python3 scripts/calculate.py` while the registry metadata declares no required binaries; python3 should be considered a runtime dependency but is not declared.

Instruction Scope

SKILL.md instructs reading local files and running the included script only (no network or unrelated file access). However it references hidden Q31/Q32 and a hidden 'DRUNK' type: Q31/Q32 exist in references/questions.md and SKILL.md describes a DRUNK trigger, but scripts/calculate.py expects exactly 30 answers (Q1–Q30) and the TYPES mapping in the script does not include a 'DRUNK' pattern — meaning the documented hidden-type behavior is not implemented by the script. This is an internal inconsistency (likely a bug) that can cause unexpected results if the agent follows the SKILL.md.

Install Mechanism

No install spec (instruction-only with included files). Low risk — nothing downloaded from external URLs and no archive extraction. Included Python script is local and executed, which is expected for this skill's purpose.

Credentials

The skill requests no environment variables, credentials, or config paths. The code does not read environment variables or make network calls. This is proportionate to a quiz/utility skill.

Persistence & Privilege

Skill does not request persistent presence (always:false) and does not modify other skills or system-wide settings. It only reads its bundled reference files and runs a local calculation script.

What to consider before installing

This skill is largely what it says: a local SBTI quiz using bundled question/type files and a Python script. Before installing: 1) Be aware the SKILL.md assumes python3 is available — the registry metadata does not declare this; ensure your environment has python3 if you want to run the script. 2) The README claims a hidden DRUNK type triggered by Q31/Q32, but the included calculate.py processes 30 answers (Q1–Q30) and the TYPES mapping lacks a DRUNK pattern — this mismatch means the hidden-type behavior is not implemented; if you expect that behavior, inspect/modify the script. 3) No network calls or secrets are requested, so there is no obvious exfiltration, but the skill will collect users' quiz answers (personal responses); consider privacy before logging or sharing results. 4) If you plan to use this autonomously, verify the script's logic and edge-case handling (input validation, hidden-question handling) to avoid surprising outputs.

算法

15维度，每维度2题（选项值1/2/3），维度得分=两题之和（2-6）。 得分→等级：≤3=L, 4=M, ≥5=H。L=1, M=2, H=3。 欧氏距离匹配27种人格，取最近者。

隐藏彩蛋：Q31选[1]喝酒 + Q32选[1]爱喝 → 强制匹配 DRUNK（酒鬼）。

维度定义

维度	名称	模型
S1	自尊自信	自我模型
S2	自我清晰度	自我模型
S3	核心价值	自我模型
E1	依恋安全感	情感模型
E2	情感投入度	情感模型
E3	边界与依赖	情感模型
A1	世界观倾向	态度模型
A2	冒险精神	态度模型
A3	目标感	态度模型
Ac1	行动力	行动模型
Ac2	果断性	行动模型
Ac3	执行力	行动模型
So1	社交主动性	社交模型
So2	亲密需求	社交模型
So3	自我呈现	社交模型

维度向量顺序：[S1, S2, S3, E1, E2, E3, A1, A2, A3, Ac1, Ac2, Ac3, So1, So2, So3]

测评流程

模式一：对话式测评（推荐）

逐题或分组（5题一组）发送给用户，收集答案后计算结果。

1. 读取 references/questions.md 获取完整题库
2. 按组发送题目（每题包含题目+3个选项）
3. 记录用户每题选择的选项值（1/2/3）
4. 收集完30题后，运行 scripts/calculate.py 计算结果
5. 读取 references/types.md 获取结果描述，输出给用户

输出格式：

🎉 你的SBTI人格是：[代码] [中文名]！
[intro语]

📊 十五维度评分：
[5个模型的雷达图或列表]

[result描述前300字]

模式二：快速查询

用户问"怎么测出XXX人格"时：

1. 读取 references/types.md 找到该类型的向量模式
2. 读取 references/questions.md
3. 反推每维度需要的目标得分，给出答题建议

答题策略：

• 目标H（≥5）：该维度两题都选3
• 目标M（=4）：该维度两题选3+1或1+3
• 目标L（≤3）：该维度两题都选1

模式三：反向查询

用户提供自己的30题答案，运行 scripts/calculate.py 直接出结果。

脚本使用

# 计算结果：传入30个答案值（逗号分隔，按Q1-Q30顺序）
python3 scripts/calculate.py "3,2,1,3,2,1,3,2,1,3,2,1,3,2,1,3,2,1,3,2,1,3,2,1,3,2,1,3,2,1"

# 输出：匹配的类型代码、名称、相似度、各维度等级

注意事项

• Q22（Ac2果断性）是隐藏题："此题没有题目，请盲选"——直接发送选项让用户选
• Q31/Q32 是酒鬼触发题，仅当用户提到或想触发隐藏类型时才询问
• 评价结果仅供娱乐，提醒用户不要当真