MbtiClaude
v1.1.0Infer MBTI personality types by analyzing user prompts across multiple AI tools (Claude Code, Codex, Gemini, OpenCode, OpenClaw) | 通过分析用户在多个 AI 工具的提示词来推测 MBT...
⭐ 0· 38·0 current·0 all-time
by@xmanrui
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The name/description (inferring MBTI from prompt history) align with the code/instructions that extract local prompt histories from multiple AI tools. Extracting those files is a reasonable capability for this purpose, but the registry metadata declares no required config paths while the runtime instructions explicitly read many home-directory history files — an inconsistency worth noting.
Instruction Scope
SKILL.md instructs the agent to read prompt/history files for Claude Code, Codex, Gemini, OpenCode, and OpenClaw (paths under the user's home). Those actions are in-scope for analyzing prompts, but they involve reading potentially sensitive user conversations and may include secrets. The instructions do not include explicit safeguards (e.g., redaction, local-only enforcement) beyond claiming local-only analysis.
Install Mechanism
The skill is instruction-only with no install spec in the registry (lowest install risk). The README suggests fetching code from GitHub or via curl/npx, but those are usage notes rather than automated install steps included in the package.
Credentials
The registry lists no required config paths or credentials, yet the SKILL.md and README rely on reading many home-directory files (~/.claude, ~/.codex, ~/.gemini, ~/.local/state/opencode, ~/.openclaw). That mismatch is an incoherence: the skill needs filesystem access to several tool histories but did not declare this. Access to these files can expose sensitive user data; the skill does not request or document any least-privilege constraints.
Persistence & Privilege
always:false and no special privileges are requested, which is appropriate. However, autonomous invocation is allowed (platform default); combined with the ability to read many local histories, this increases the potential blast radius if you enable/allow the skill without review.
What to consider before installing
This skill will look through many local AI history files in your home directory to infer MBTI. Before installing, (1) inspect the SKILL.md/README on the GitHub repo to confirm nothing uploads data or calls external endpoints, (2) review the actual files it will read (~/.claude, ~/.codex, ~/.gemini, ~/.local/state/opencode, ~/.openclaw) to see whether they contain sensitive content you don't want analyzed, (3) prefer running it offline or in an isolated account/container, and (4) ask the author to declare the required config paths and add explicit local-only redaction/sanitization if you plan to use it with real conversations. If you are uncomfortable exposing local conversation history, do not install or run this skill.Like a lobster shell, security has layers — review code before you run it.
latest
MBTIClaude
Infer MBTI personality types by analyzing user prompts across multiple AI tools (Claude Code, Codex, Gemini, OpenCode, OpenClaw).
通过分析用户在多个 AI 工具（Claude Code、Codex、Gemini、OpenCode、OpenClaw）中的提示词来推测 MBTI 性格类型。
Features | 功能
- Automatically extract user prompt history from five AI tools | 自动提取五种 AI 工具的用户提示词历史
- Analyze communication style, information processing, decision-making patterns, and lifestyle | 分析沟通风格、信息处理方式、决策模式和生活方式
- Infer MBTI type based on behavioral patterns | 基于行为模式推测 MBTI 类型
- Provide detailed feature analysis and recommendations | 提供详细的特征分析和建议
Usage | 使用方法
/mbticlaude
Analysis Dimensions | 分析维度
1. Communication Style (E vs I) | 沟通风格
- Prompt length and detail level | 提示词长度和详细程度
- Social expression vs task-oriented | 社交性表达 vs 任务导向
- Richness of background information | 背景说明的丰富程度
2. Information Processing (S vs N) | 信息处理
- Focus on concrete details vs abstract concepts | 关注具体细节 vs 抽象概念
- Implementation-oriented vs concept-oriented | 实现导向 vs 概念导向
- Expression of creative thinking | 创造性思维的体现
3. Decision Making (T vs F) | 决策方式
- Logical vs emotional | 逻辑性 vs 情感性
- Task-oriented vs interpersonal considerations | 任务导向 vs 人际关系考虑
- Technical language vs emotional expression | 技术语言 vs 情感表达
4. Lifestyle (J vs P) | 生活方式
- Planning vs flexibility | 计划性 vs 灵活性
- Structured vs exploratory | 结构化 vs 探索性
- Rapid iteration vs systematic planning | 快速迭代 vs 系统规划
Data Sources | 数据来源
- Claude Code:
~/.claude/history.jsonl - Codex:
~/.codex/history.jsonl - Gemini:
~/.gemini/tmp/*/chats/*.json - OpenCode:
~/.local/state/opencode/prompt-history.jsonl - OpenClaw:
~/.openclaw/agents/*/sessions/*.jsonl
Important Notes | 注意事项
⚠️ Important Reminders | 重要提醒：
- Sample based on work scenarios, may not reflect true personality | 样本基于工作场景，可能不反映真实性格
- MBTI itself has scientific controversies, for reference only | MBTI 本身科学性存疑，仅供参考
- This is behavioral pattern inference, not professional psychological assessment | 这是基于行为模式的推测，非专业心理测评
- Requires sufficient historical data for accurate conclusions | 需要足够的历史数据才能得出较准确的结论
Implementation | 实现
#!/usr/bin/env python3
import json
import os
from pathlib import Path
from collections import Counter
def extract_claude_code_prompts(limit=100):
"""提取 Claude Code 用户提示词"""
history_file = Path.home() / ".claude" / "history.jsonl"
prompts = []
if history_file.exists():
with open(history_file, 'r') as f:
for line in f:
try:
data = json.loads(line)
if data.get('display'):
prompts.append(data['display'])
except:
continue
return prompts[-limit:] if prompts else []
def extract_codex_prompts(limit=100):
"""提取 Codex 用户提示词"""
history_file = Path.home() / ".codex" / "history.jsonl"
prompts = []
if history_file.exists():
with open(history_file, 'r') as f:
for line in f:
try:
data = json.loads(line)
if data.get('text'):
prompts.append(data['text'])
except:
continue
return prompts[-limit:] if prompts else []
def extract_gemini_prompts(limit=100):
"""提取 Gemini 用户提示词"""
gemini_dir = Path.home() / ".gemini" / "tmp"
prompts = []
if gemini_dir.exists():
for session_file in gemini_dir.rglob("session-*.json"):
try:
with open(session_file, 'r') as f:
data = json.load(f)
messages = data.get('messages', [])
for msg in messages:
if msg.get('role') == 'user':
content = msg.get('content', '')
if content:
prompts.append(content)
except:
continue
return prompts[-limit:] if prompts else []
def extract_opencode_prompts(limit=100):
"""提取 OpenCode 用户提示词"""
history_file = Path.home() / ".local" / "state" / "opencode" / "prompt-history.jsonl"
prompts = []
if history_file.exists():
with open(history_file, 'r') as f:
for line in f:
try:
data = json.loads(line)
if data.get('input'):
prompts.append(data['input'])
except:
continue
return prompts[-limit:] if prompts else []
def extract_openclaw_prompts(limit=100):
"""提取 OpenClaw 用户提示词"""
agents_dir = Path.home() / ".openclaw" / "agents"
prompts = []
if agents_dir.exists():
for session_file in agents_dir.rglob("*.jsonl"):
try:
with open(session_file, 'r') as f:
for line in f:
try:
data = json.loads(line)
if data.get('type') == 'message' and data.get('message', {}).get('role') == 'user':
content = data['message'].get('content', [])
for item in content:
if item.get('type') == 'text':
text = item.get('text', '')
# 过滤系统消息
if not any(x in text for x in ['Sender (untrusted', 'Conversation info',
'Health check', 'Queued', 'Continue where',
'subagent task', 'Pre-compaction']):
prompts.append(text)
except:
continue
except:
continue
return prompts[-limit:] if prompts else []
def analyze_communication_style(prompts):
"""分析沟通风格 (E vs I) - 优化版"""
if not prompts:
return "I"
avg_length = sum(len(p) for p in prompts) / len(prompts)
short_prompts = sum(1 for p in prompts if len(p) < 15) # 降低阈值从20到15
short_ratio = short_prompts / len(prompts)
# 计算详细度得分
detailed_prompts = sum(1 for p in prompts if len(p) > 100)
detailed_ratio = detailed_prompts / len(prompts)
# 检查是否有问号（提问倾向）
question_count = sum(1 for p in prompts if '?' in p or '？' in p or 'how' in p.lower() or 'why' in p.lower() or '如何' in p or '为什么' in p)
question_ratio = question_count / len(prompts)
# 综合判断：提高 E 型出现概率
e_score = 0
if avg_length > 40: # 降低阈值从50到40
e_score += 1
if short_ratio < 0.4: # 提高阈值从0.3到0.4
e_score += 1
if detailed_ratio > 0.15: # 有15%以上的详细提示词
e_score += 1
if question_ratio > 0.2: # 有20%以上的提问
e_score += 1
return "E" if e_score >= 2 else "I"
def analyze_information_processing(prompts):
"""分析信息处理 (S vs N) - 优化版，支持英文"""
if not prompts:
return "S"
# 扩展关键词，增加英文支持和权重
abstract_keywords = {
# 中文（权重2）
'概念': 2, '系统': 2, '架构': 3, '方案': 2, '思路': 2, '抽象': 3,
'模式': 2, '框架': 2, '理念': 2, '愿景': 2, '战略': 2,
# 英文（权重2）
'concept': 2, 'system': 2, 'architecture': 3, 'solution': 2, 'approach': 2,
'pattern': 2, 'framework': 2, 'strategy': 2, 'vision': 2, 'abstract': 3
}
concrete_keywords = {
# 中文（权重1）
'具体': 1, '实现': 1, '细节': 1, '步骤': 1, '操作': 1,
'按钮': 1, '颜色': 1, '间距': 1, '像素': 1, '字体': 1,
# 英文（权重1）
'specific': 1, 'implement': 1, 'detail': 1, 'step': 1, 'operation': 1,
'button': 1, 'color': 1, 'spacing': 1, 'pixel': 1, 'font': 1
}
abstract_score = sum(weight for p in prompts for kw, weight in abstract_keywords.items() if kw in p.lower())
concrete_score = sum(weight for p in prompts for kw, weight in concrete_keywords.items() if kw in p.lower())
# 降低 N 型判定门槛
return "N" if abstract_score > concrete_score * 0.8 else "S"
def analyze_decision_making(prompts):
"""分析决策方式 (T vs F) - 优化版，支持英文"""
if not prompts:
return "T"
# 扩展关键词，增加英文支持
logical_keywords = [
# 中文
'优化', '效率', '性能', '统计', '分析', '测试', '修复', '实现',
'算法', '数据', '指标', '评估', '验证', '检查', '对比',
# 英文
'optimize', 'efficiency', 'performance', 'statistics', 'analyze', 'test', 'fix', 'implement',
'algorithm', 'data', 'metric', 'evaluate', 'verify', 'check', 'compare'
]
emotional_keywords = [
# 中文
'感觉', '喜欢', '希望', '想要', '谢谢', '辛苦', '帮忙', '麻烦',
'期待', '担心', '开心', '满意', '友好', '舒服', '美观',
# 英文
'feel', 'like', 'hope', 'want', 'thanks', 'thank', 'please', 'appreciate',
'expect', 'worry', 'happy', 'satisfied', 'friendly', 'comfortable', 'beautiful'
]
logical_count = sum(1 for p in prompts for kw in logical_keywords if kw in p.lower())
emotional_count = sum(1 for p in prompts for kw in emotional_keywords if kw in p.lower())
# 降低 T 型判定门槛，让 F 型更容易出现
return "T" if logical_count > emotional_count * 1.5 else "F"
def analyze_lifestyle(prompts):
"""分析生活方式 (J vs P) - 优化版，支持英文"""
if not prompts:
return "P"
# 扩展关键词，增加英文支持
planning_keywords = [
# 中文
'计划', '方案', '设计', 'PRD', '文档', '规划', '安排', '流程',
'步骤', '阶段', '里程碑', '时间表', 'roadmap',
# 英文
'plan', 'design', 'schedule', 'organize', 'structure', 'process',
'procedure', 'milestone', 'timeline', 'roadmap', 'workflow'
]
flexible_keywords = [
# 中文
'试试', '测试', '优化', '调整', '改', '换', '尝试', '探索',
'实验', '迭代', '看看', '试一下', '改进',
# 英文
'try', 'test', 'experiment', 'adjust', 'change', 'modify', 'explore',
'iterate', 'adapt', 'improve', 'refine', 'tweak'
]
planning_count = sum(1 for p in prompts for kw in planning_keywords if kw in p.lower())
flexible_count = sum(1 for p in prompts for kw in flexible_keywords if kw in p.lower())
# 保持原有逻辑
return "P" if flexible_count > planning_count else "J"
def analyze_mbti():
"""主函数：分析 MBTI"""
print("🔍 正在提取用户提示词...")
# 提取所有工具的提示词
claude_prompts = extract_claude_code_prompts()
codex_prompts = extract_codex_prompts()
gemini_prompts = extract_gemini_prompts()
opencode_prompts = extract_opencode_prompts()
openclaw_prompts = extract_openclaw_prompts()
# 统计
print(f"\n📊 数据统计：")
print(f" Claude Code: {len(claude_prompts)} 条")
print(f" Codex: {len(codex_prompts)} 条")
print(f" Gemini: {len(gemini_prompts)} 条")
print(f" OpenCode: {len(opencode_prompts)} 条")
print(f" OpenClaw: {len(openclaw_prompts)} 条")
all_prompts = claude_prompts + codex_prompts + gemini_prompts + opencode_prompts + openclaw_prompts
total = len(all_prompts)
print(f" 总计: {total} 条\n")
if total < 10:
print("⚠️ 数据量太少，无法进行准确分析")
return
# 分析四个维度
print("🧠 正在分析 MBTI 维度...\n")
e_or_i = analyze_communication_style(all_prompts)
s_or_n = analyze_information_processing(all_prompts)
t_or_f = analyze_decision_making(all_prompts)
j_or_p = analyze_lifestyle(all_prompts)
mbti_type = f"{e_or_i}{s_or_n}{t_or_f}{j_or_p}"
# 输出结果
print("=" * 50)
print(f"🎯 你的 MBTI 类型：{mbti_type}")
print("=" * 50)
mbti_names = {
"INTJ": "建筑师", "INTP": "逻辑学家", "ENTJ": "指挥官", "ENTP": "辩论家",
"INFJ": "提倡者", "INFP": "调停者", "ENFJ": "主人公", "ENFP": "竞选者",
"ISTJ": "物流师", "ISFJ": "守卫者", "ESTJ": "总经理", "ESFJ": "执政官",
"ISTP": "鉴赏家", "ISFP": "探险家", "ESTP": "企业家", "ESFP": "表演者"
}
print(f"\n类型名称：{mbti_names.get(mbti_type, '未知')}\n")
print("📋 维度分析：")
print(f" {e_or_i} - {'内向' if e_or_i == 'I' else '外向'}：简洁高效沟通" if e_or_i == 'I' else f" {e_or_i} - 外向：详细表达")
print(f" {s_or_n} - {'直觉' if s_or_n == 'N' else '感觉'}：{'抽象思维，系统化方案' if s_or_n == 'N' else '关注具体细节'}")
print(f" {t_or_f} - {'思考' if t_or_f == 'T' else '情感'}：{'纯理性，关注效率' if t_or_f == 'T' else '考虑情感因素'}")
print(f" {j_or_p} - {'判断' if j_or_p == 'J' else '知觉'}：{'计划性强' if j_or_p == 'J' else '灵活探索，快速迭代'}")
print("\n⚠️ 重要提醒：")
print(" 1. 基于工作场景的行为推测，可能不反映真实性格")
print(" 2. MBTI 本身科学性存疑，仅供参考")
print(" 3. 这是基于行为模式的推测，非专业心理测评")
if __name__ == "__main__":
analyze_mbti()
示例输出
🔍 正在提取用户提示词...
📊 数据统计：
Claude Code: 100 条
Codex: 100 条
Gemini: 62 条
OpenCode: 2 条
OpenClaw: 100 条
总计: 364 条
🧠 正在分析 MBTI 维度...
==================================================
🎯 你的 MBTI 类型：INTP
==================================================
类型名称：逻辑学家
📋 维度分析：
I - 内向：简洁高效沟通
N - 直觉：抽象思维，系统化方案
T - 思考：纯理性，关注效率
P - 知觉：灵活探索，快速迭代
⚠️ 重要提醒：
1. 基于工作场景的行为推测，可能不反映真实性格
2. MBTI 本身科学性存疑，仅供参考
3. 这是基于行为模式的推测，非专业心理测评
许可证
MIT
Comments
Loading comments...