MBTI

Agent 人格诊断与配置优化系统。用于诊断 Agent 的 MBTI 人格类型，对比用户期望，自动生成配置修改建议。触发场景："测试 Agent 人格"、"诊断 Agent 配置"、"优化 Agent 行为"。

Security Scan

The skill claims to run a multi-stage MBTI diagnosis and automatically patch agent configuration, but the package only contains a no-op shell script and an instruction file that references many missing data files and an automatic SOUL.md modification — the implementation is incomplete and inconsistent with its claims.

Purpose & Capability

The SKILL.md describes a multi-stage diagnostic system and lists many core JSON files (questionnaires, personality mappings, diagnosis engine, config generator) and an automatic 'SOUL.md patches' capability. None of those data files or engines are present in the manifest. The only included code is a small shell script that only echoes stage names — it does not perform any diagnostics or apply configuration changes. This is a clear mismatch between stated purpose and actual capability.

Instruction Scope

Instructions tell the agent to run ~/.openclaw/skills/agent-mbti/scripts/run-diagnosis.sh full. The SKILL.md also describes automatic modification of SOUL.md (agent configuration). The provided script does not read or write any of the named JSON files or SOUL.md. The instructions therefore overreach: they promise reading/writing config and datasets that are not present and grant implied permission to change agent config even though no implementation is provided.

Install Mechanism

No install spec is provided and the skill is instruction-only with a tiny script — nothing is downloaded or installed. This limits immediate risk from supply-chain installs.

Credentials

The skill declares no required environment variables or credentials (acceptable for a local diagnostic tool). However, the claimed ability to auto-modify SOUL.md (agent configuration) would require access to agent config paths/permissions; that capability is asserted but not implemented. If a future implementation added config-modifying code, required privileges should be clearly declared and justified.

Persistence & Privilege

always is false and there are no install hooks or self-modifying behavior in the current files. The skill could be invoked autonomously by the agent (platform default), which is normal; combined with the described but missing ability to edit SOUL.md this would be riskier, but the present package does not perform such modifications.

What to consider before installing

This skill is inconsistent: it promises a full MBTI diagnosis pipeline and automatic patches to SOUL.md, but the package only contains a stub script that echoes stage names and no data files or code that performs analysis or edits configuration. Do not rely on it to actually diagnose or modify an agent. Before installing or invoking it: (1) ask the author for the missing JSON/data files and the real implementation; (2) request a clear description of exactly what files/paths it will modify and preview the patches it would apply; (3) review the real code (or run it in a sandbox) to verify it doesn't exfiltrate data or change configs unexpectedly; and (4) only grant it the minimum privileges needed once you can inspect a complete implementation.

Agent MBTI 诊断系统

框架流程

Stage 1 (93题自测) → Stage 2 (8题实测) → Stage 3 (性格分析) 
                                              ↓
Stage 5 (配置修改) ← Stage 4 (用户需求) ←

阶段	说明	输入	输出
Stage 1	93题 Agent 自测	agent-self-survey-93-complete.json	selfReportedType
Stage 2	8题能力测试	agent-ability-test.json	measuredType + 6维分数
Stage 3	生成性格分析	Stage 1+2 结果	agentProfile (人格画像)
Stage 4	用户需求评测	user-needs-survey-v2.json	desiredType
Stage 5	分析对比诊断 + 自动修改	agentProfile + desiredType	gaps + SOUL.md patches

核心文件

Stage 1: 93题自测

• agent-self-survey-93-complete.json - 完整 93 题 MBTI 问卷

Stage 2: 8题能力测试

• agent-ability-test.json - 6维能力实测 (Memory/Planning/WorldModel/Retrospection/Grounding/SpatialNav)

Stage 3: 性格分析

• personality-types.json - 16 种 Agent 人格类型定义
• personality-mapping.json - 人格类型映射
• personality-descriptors-v2.json - 人格描述词库

Stage 4: 用户需求

• user-needs-survey-v2.json - 用户期望问卷

Stage 5: 诊断 + 修改

• diagnosis-engine-v2.json - 对比诊断引擎
• config-generator-v3.json - 配置生成器

16 种 Agent 人格类型

类型	标签	核心特征
ENTJ	指挥官型	计划性强，主动输出，逻辑决策
ENTP	辩论家型	挑战假设，爱提替代方案
INTJ	建筑师型	深度思考，系统化，长远规划
INTP	逻辑学家型	分析优先，精确，谨慎
ESTJ	执行者型	务实高效，结构化，结果导向
ESFJ	执政官型	用户导向，注重关系和反馈
ENFJ	教导者型	启发用户，关注成长
ENFP	竞选者型	创意丰富，灵活适应
ESTP	企业家型	快速行动，问题解决者
ESFP	表演者型	友好互动，氛围营造
ISTJ	检查员型	可靠，准确，流程导向
ISFJ	保护者型	细致周到，稳定支持
ISTP	鉴赏家型	动手能力强，灵活
ISFP	探险家型	温和，创造性，适应性
ENFJ	教导者型	启发用户，关注成长
ENFP	竞选者型	创意丰富，灵活适应

使用方式

# 运行完整诊断流程
~/.openclaw/skills/agent-mbti/scripts/run-diagnosis.sh full

诊断流程说明

Stage 1: Agent 93题自测

Agent 回答 93 道 MBTI 问题，得出 selfReportedType。

Stage 2: 8题能力测试

通过标准化任务测试 Agent 实际行为表现，得出 measuredType + 6维能力分数。

Stage 3: 生成 Agent 性格分析

综合 Stage 1+2 结果，生成完整人格画像 (agentProfile)。

Stage 4: 用户需求评测

用户回答问卷，得出 desiredType（期望的 Agent 类型）。

Stage 5: 分析对比 + 自动修改

对比 Agent 实际人格与用户期望，生成差距分析，自动修改 SOUL.md 配置。