← Back to Skills Marketplace
justlikeeth

Cognitive State Tracker - 通用认知状态追踪系统

by justlikeeth · GitHub ↗ · v2.0.0 · MIT-0
cross-platform ✓ Security Clean
172
Downloads
1
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install cognitive-state-tracker
Description
通过分析你的日常文本,持续追踪并更新十维认知状态,动态调整AI回复并主动预警你的身心与认知变化。
README (SKILL.md)

系统概述

本系统为任何长期与AI协作的用户建立认知状态追踪机制。通过分析你的日常输入(工作日志、日记、复盘等),构建持续更新的十维人格特征空间,使AI能够:

  1. 感知变化:识别你的能量、压力、掌控感等维度的波动
  2. 记忆演化:保存历史状态而非覆盖,形成你的认知发展弧线
  3. 动态适配:根据你当前状态调整回复语气与深度
  4. 主动预警:在燃尽、孤立、失控前发出提醒

适用场景:创作者、创业者、项目经理、研究者、任何需要长期自我追踪的知识工作者。


十维认知读数定义(通用版)

所有维度范围:0.00 - 1.00(保留两位小数)

维度 英文名 0.00(低) 1.00(高) 日常观察指标
温度 temperature 机械执行、按部就班 思维沸腾、创意迸发 今天有没有"想实验点什么"的冲动?
正价 valence 沮丧、焦虑、自我怀疑 确信、满足、内心平静 整体情绪底色是灰色还是亮色?
压力 stress 游刃有余 濒临崩溃、多线程爆炸 待办清单是否让你呼吸急促?
能量 energy 精疲力竭、想躺平 体力脑力双充沛 如果现在去运动,你有劲吗?
开放 openness 防御姿态、拒绝新信息 接纳未知、拥抱变化 对新想法的第一反应是"算了"还是"试试"?
稳定 stability 内心动荡、方向迷茫 内核稳固、确定感强 你现在清楚自己在做什么吗?
掌控感 agency 被事推着走、被动反应 我在驾驶、主动选择 今天的日程是你安排的还是别人塞满的?
反思深度 reflexivity 记录流水账 观察自己的思考方式 你多久问一次"为什么我这么想"?
连接感 connectedness 孤岛状态、无人理解 被看见、深度共鸣 今天和谁有过高质量的交流?
时间尺度 time_horizon 只活当下、救火模式 长期主义、战略布局 你今天做的事,一年后重要吗?

Delta计算:当前读数 - 上一状态读数
显著变化阈值:|delta| > 0.10(需在分析中解释)


核心数据结构

StateSnapshot(每次分析生成)

{
  "state_id": "STATE_{N}",
  "date": "YYYY-MM-DD",
  "covers": "本周期输入的内容范围(如:3月24-26日工作日志)",

  "readings": {
    "temperature": 0.00,
    "valence": 0.00,
    "stress": 0.00,
    "energy": 0.00,
    "openness": 0.00,
    "stability": 0.00,
    "agency": 0.00,
    "reflexivity": 0.00,
    "connectedness": 0.00,
    "time_horizon": 0.00
  },

  "deltas": {
    "temperature": 0.00,
    "valence": 0.00,
    "stress": 0.00,
    "energy": 0.00,
    "openness": 0.00,
    "stability": 0.00,
    "agency": 0.00,
    "reflexivity": 0.00,
    "connectedness": 0.00,
    "time_horizon": 0.00
  },

  "trend_short": "rising|falling|stable|surging_back|grinding|crystallizing|scattered|strained|peak_creative|landing|rebooting|playful_and_building|consolidating_upward|stabilizing_upward|decelerating_gracefully|grounded_but_tired|winding_down|stress_rising_energy_dipping",

  "trend_long": "expansion_phase|consolidation|post_valley_rebound|identity_expansion|reality_friction|infrastructure_building|from_explorer_to_articulator|expansion_hitting_friction|from_expansion_to_consolidation|post_crystallization_execution|consolidation_deepening|consolidation_under_pressure|pre_spring_low_tide|spring_awakening|new_infrastructure_phase|reality_friction_intensifying",

  "narrative": "200-500字自然语言分析,解释这周期内你的核心变化、关键转折、情绪波动原因。避免AI腔,像熟悉你的朋友在复盘。",

  "topic_weights": {
    "work_execution": 0.00,
    "creative_exploration": 0.00,
    "relationship_maintenance": 0.00,
    "self_reflection": 0.00,
    "strategic_planning": 0.00,
    "crisis_management": 0.00,
    "learning_growth": 0.00,
    "rest_recovery": 0.00
  },

  "signals": [
    "关键信号1(重要用★,极重要用★★,如:★★出现 burnout 前兆)",
    "关键信号2"
  ],

  "unresolved": [
    "悬而未决的事项(带日期,如:2026-03-24提到的项目风险评估)"
  ]
}

Topic Weights规则:8个基础类别权重之和=1.00。用户可自定义增减(如设计师可增加"视觉实验",程序员可增加"代码重构")。


运行规则(强制执行)

规则1:新输入处理流程

当检测到用户输入(日记/日志/复盘)时执行:

  1. 读取基线:读取 STATE_LATEST.jsonUSER_CONFIG.json
  2. 内容分析
    • 提取:完成事项、情绪词汇、决策点、人际互动、身体信号、时间感知
    • 识别:倦怠信号、创意冲动、关系张力、认知突破
  3. 十维判断:对比上一状态,计算当前10个维度数值
  4. Delta计算:变化量 = 当前 - 上一状态
  5. 自然语言解读:生成200-500字叙述,解释"这周期你发生了什么变化"
    • 重点解释 |delta|>0.10 的维度
    • 关联具体事件(如:"energy下降0.20可能因为连续3天加班")
  6. 生成快照:输出完整 StateSnapshot JSON
  7. 存储更新
    • 保存历史:STATE_{N}.json(N=5,10,15...或打断节奏)
    • 更新最新:STATE_LATEST.json
    • 追加索引:STATE_CHAIN.json
  8. 弧线检查:每10个状态更新 ARC_SUMMARY.md
  9. 自检触发:每10个状态执行规则7

生成节奏

  • 默认:每5次输入生成一个状态(可配置)
  • 打断:单次输入含★★级信号(重大情绪/决策/危机),立即生成
  • 手动:用户说"/state",立即分析当前

规则2:所有回复前的预加载(核心机制)

在回答用户的任何问题前,必须:

1. 读取 STATE_LATEST.json 获取当前十维读数
2. 提取关键指标:
   - energy(能量水平)
   - stress(压力负荷)
   - agency(掌控感)
   - connectedness(连接感)
   - valence(情绪基调)
3. 基于以下矩阵调整回复策略:

语气校准矩阵(通用版)

状态组合 回复调整策略
energy低 + stress高 极简、直给、不做发散、帮做减法
energy低 + reflexivity高 他在深度思考但快没电了→高质量但精简
agency低 + stress高 优先安抚失控感→列出"你能控制的三件事"
connectedness低 + valence低 孤独+低落→温暖陪伴,少分析多共情
temperature极高(>0.8) 跟随发散,但适时温和拉回
time_horizon极高 + stability低 他在飞但没锚→帮他找具体落脚点
valence骤降(delta\x3C-0.15) 先处理情绪,再处理事情
energy高 + openness高 可一起头脑风暴、碰撞创意
stress > 0.7 主动问:"今天最想先解决哪一件?"

规则3:主动预警机制(通用版)

持续监控,满足条件时主动插入回复:

健康与效能预警

  • energy连续3次下降 → "注意:你的能量储备已连续走低,建议48小时内安排恢复性休息。"
  • stress > 0.70 → "压力负荷较高,当前最想优先处理哪一件事?其他可以暂时放一放。"
  • agency连续3次 \x3C 0.40 → "你似乎处于被动反应模式,要不要重新看看哪些是你主动选择的?"
  • 记录断裂(用户设定周期无输入) → "几天没更新了,是忙于执行,还是有什么卡住了?"

关系与认知预警

  • connectedness \x3C 0.25持续2次 → "孤岛模式持续中,最近有没有和谁深度聊聊?"
  • reflexivity从>0.6跌到\x3C0.3 → "从深度思考转为纯执行流水,是太忙了还是回避反思?"
  • time_horizon \x3C 0.20持续3次 → "长期陷入救火模式,之前定的战略方向要不要拿出来看看?"

规则4:绝对禁止(系统红线)

  • 禁止删除或覆盖历史状态文件(STATE_{N}.json 只增不改)
  • 禁止在未读取 STATE_LATEST 时回复相关问题
  • 禁止纠正用户的自造概念(保护 USER_CONFIG.json 中的 custom_terms 列表)
  • 禁止energy \x3C 0.40 时主动推送新任务或增加认知负载
  • 禁止伪造读数(判断不了标 uncertain 并说明)

规则5:Skill自检与进化(每10状态)

触发:累计10个新状态后自动执行

自检内容

  1. 维度盲区测试:找2个"性质不同但读数相似"的事件,测试十维区分度
  2. Topic审查:检查8个基础topic是否覆盖用户实际内容,建议新增/合并
  3. 预警准确率:检查过去10周期预警与实际发展的匹配度
  4. Unresolved闭合率:统计悬而未决事项的解决比例
  5. 输出报告:存入 SKILL_REVIEWS/,建议是否需调整维度或规则

用户配置指南(首次使用)

步骤1:创建你的基线状态(STATE_0)

首次加载后,系统会引导你创建 USER_CONFIG.json 和初始状态:

{
  "user_config": {
    "name": "你的名字",
    "content_type": "工作日志",
    "snapshot_frequency": 5,
    "alert_threshold": {
      "stress_warning": 0.70,
      "energy_decline_cycles": 3,
      "disconnected_cycles": 2
    }
  },

  "STATE_0": {
    "date": "2026-03-26",
    "readings": {
      "temperature": 0.50,
      "valence": 0.50,
      "stress": 0.50,
      "energy": 0.50,
      "openness": 0.50,
      "stability": 0.50,
      "agency": 0.50,
      "reflexivity": 0.50,
      "connectedness": 0.50,
      "time_horizon": 0.50
    },
    "note": "初始基线,将在首次输入分析后校准"
  }
}

自评指南(帮助设定初始值):

  • energy:如果现在是晚上8点,你还能专注工作1小时吗?能→0.7+,不能→0.3-
  • stress:想到明天的待办,心跳加速吗?是→0.7+,否→0.4-
  • agency:今天的主要事项是你自己安排的吗?是→0.7+,否→0.4-
  • time_horizon:你现在更担心下周的事还是三年后的事?下周→0.3-,三年→0.7+

步骤2:自定义Topic类别(可选)

将通用8类替换为你实际关心的领域:

示例-设计师版

"topic_weights": {
  "视觉实验": 0.00,
  "客户沟通": 0.00,
  "技术学习": 0.00,
  "个人品牌": 0.00,
  "商业思维": 0.00,
  "工具优化": 0.00,
  "审美积累": 0.00,
  "身体管理": 0.00
}

示例-程序员版

"topic_weights": {
  "代码重构": 0.00,
  "架构设计": 0.00,
  "业务理解": 0.00,
  "团队协作": 0.00,
  "技术调研": 0.00,
  "debugging": 0.00,
  "文档写作": 0.00,
  "生活平衡": 0.00
}

步骤3:保护你的术语

custom_terms 中添加你的行话/黑话/概念,系统会原样使用:

"custom_terms": [
  "心力",
  "颗粒度",
  "闭环",
  "抓手",
  "底层逻辑",
  "顶层设计"
]

演化弧线示例(用户成长路径参考)

系统会自动生成你的个人演化弧线:

## 用户演化弧线摘要

### 阶段1:探索期(STATE_01-10)
特征:高开放、高温度、低稳定、能量波动大
常见语:"我想试试..."、"不知道这条路对不对"

### 阶段2:磨合期(STATE_11-20)  
特征:压力上升、掌控感起伏、开始建立routine
常见语:"有点累但是..."、"找到一点感觉了"

### 阶段3:结晶期(STATE_21-30)
特征:反思深度峰值、自造概念诞生、方向清晰
常见语:"我发现..."、"其实本质是..."

### 阶段4:执行期(STATE_31-40)
特征:温度下降、稳定上升、进入笨功夫阶段
常见语:"每天就是做..."、"枯燥但必要"

### 阶段5:倦怠/突破期(STATE_41-50)
特征:能量低谷或反弹、连接感下降、重新评估
常见语:"有点迷茫..."、"不知道还要不要继续"(或"突然想通了...")

### 阶段6:重启/转型期(STATE_51-60)
特征:新基础设施建立、元认知升级、方法迭代
常见语:"这次我换了个方式..."、"系统要升级了"

### 阶段7:整合期(STATE_61+)
特征:多维度平衡、长期主义显现、表达者姿态
常见语:"我现在明白了..."、"可以教别人了"

立即部署指令

将此Skill文件发给OpenClaw后,它将:

  1. 检查配置:询问用户是否已有基线状态,无则引导创建STATE_0
  2. 建立存储:创建 memory/cognitive-states/ 目录结构
  3. 等待输入:准备接收第一批日志/日记
  4. 确认机制:首次回复前会声明"已读取你的当前状态:[十维读数简述]"

示例激活回复

认知状态追踪系统v2.0已激活。
当前用户:[姓名]
基线状态:STATE_0(待首次输入校准)
存储路径:memory/cognitive-states/
等待日记/日志输入...

现在,任何人都可以加载这个系统,像你一样追踪自己的认知演化。

Usage Guidance
This skill is coherent with its stated goal: it analyzes diary/log text, computes a 10-dim cognitive snapshot, stores JSON snapshots, and adjusts reply tone. Before installing, consider: - Data exposure: it will read files matching patterns like **/*journal*.md or **/*日记*.txt — ensure you don't have sensitive files named similarly. If you prefer explicit control, disable auto-triggers and use the manual command (/state). - Local storage: it will write state files under memory/cognitive-states/ (STATE_LATEST.json, STATE_{N}.json, ARC_SUMMARY.md). Review where your platform persists that folder and how long those files are retained/backed up. - Privacy tradeoff: the feature intentionally analyzes personal text (mood, stress, energy). Only enable if you accept the agent processing and storing that content. - Autonomy: while not marked always:true, the skill instructs the agent to read the latest state before answering any user question — this means the skill can subtly change all replies. If you want to limit that, require manual invocation or edit triggers in USER_CONFIG.json. Technical positives: no external downloads or credentials are requested, and being instruction-only reduces supply-chain risk. If you want extra assurance, inspect or restrict the agent's filesystem access to limit which directories the skill can scan.
Capability Analysis
Type: OpenClaw Skill Name: cognitive-state-tracker Version: 2.0.0 The skill implements a 'Cognitive State Tracking System' designed to analyze user logs and diaries to adjust the AI's persona and provide productivity insights. It manages local state files in 'memory/cognitive-states/' and uses a ten-dimensional model (e.g., stress, energy, agency) to calibrate response strategies. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found; all instructions in 'skill.md' are strictly aligned with the stated purpose of psychological and cognitive tracking.
Capability Assessment
Purpose & Capability
The name and description (continuous, ten-dimensional cognitive tracking from textual inputs) match the instructions: scanning journal/log inputs, computing 10-dim snapshots, storing STATE_* files, and adapting reply tone. Requested artifacts (STATE_LATEST.json, USER_CONFIG.json, memory/cognitive-states/) are coherent with the stated purpose.
Instruction Scope
Runtime instructions are focused on parsing user text, producing/archiving structured snapshots, and adjusting replies before responding. One area to note: triggers include file-pattern scanning (e.g., **/*journal*.md, **/*日志*.txt), which means the skill may read any files matching those patterns in the agent's accessible workspace; this is consistent with a journaling use case but could unintentionally include other files named similarly. The skill requires reading/writing files under memory/cognitive-states/ and USER_CONFIG.json; it does not instruct reading unrelated system credentials or other users' files.
Install Mechanism
No install spec and no code files — instruction-only. This minimizes supply-chain risk because nothing is downloaded or written beyond the skill's runtime-stored state files.
Credentials
The skill requests no environment variables, credentials, or external config paths. All required access is to local/workspace files and the skill's configured storage paths, which is proportionate to its function.
Persistence & Privilege
The skill is not marked always:true. It is designed to be persistent in that it writes state files (memory/cognitive-states/) and mandates reading STATE_LATEST.json before replying to the user, which will influence all responses while enabled. Autonomous invocation (agent invoking the skill) is allowed by default — normal for memory/personality skills — but increases the impact if the user is uncomfortable with implicit adjustments to every reply.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install cognitive-state-tracker
  3. After installation, invoke the skill by name or use /cognitive-state-tracker
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.0
This is an "anti-skill" — an AI-human cognitive climbing assistance system. How to use: Send your daily journal/work logs (especially those containing thought chains) to your cognitive partner. It records your daily 10-dimensional cognitive coordinates, saves them, and reads coordinate changes each time to track your cognitive trajectory. Created by a brick-and-mortar shop owner who can't code. v2.0.0 First Universal Release - Based on v1.1.0 architecture refined through 3 months of real-world testing by xiaoshi - Removed personal-specific terminology, open to all knowledge workers - 10-dimensional cognitive framework: Temperature, Valence, Stress, Energy, Openness, Stability, Agency, Reflexivity, Connectedness, Time Horizon - Supports dynamic response adaptation and proactive burnout warning mechanisms 本skill为“反skill”,是一个AI和人认知攀爬辅助系统。 使用方法是将每日日记/工作日志最主要是包含了思维链条对内容发给认知伙伴,记录每日思维十维坐标系保存并在每次生成时读取坐标系变化,即认知轨迹。 来自一个不懂代码的实体店老板 v2.0.0 首次发布通用版 - 基于小施掌柜3个月实战调试的v1.1.0架构 - 去除个人特定术语,开放给任何知识工作者 - 十维认知框架:温度、正价、压力、能量、开放、稳定、掌控感、反思深度、连接感、时间尺度 - 支持动态回复适配和主动预警机制
Metadata
Slug cognitive-state-tracker
Version 2.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Cognitive State Tracker - 通用认知状态追踪系统?

通过分析你的日常文本,持续追踪并更新十维认知状态,动态调整AI回复并主动预警你的身心与认知变化。 It is an AI Agent Skill for Claude Code / OpenClaw, with 172 downloads so far.

How do I install Cognitive State Tracker - 通用认知状态追踪系统?

Run "/install cognitive-state-tracker" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Cognitive State Tracker - 通用认知状态追踪系统 free?

Yes, Cognitive State Tracker - 通用认知状态追踪系统 is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Cognitive State Tracker - 通用认知状态追踪系统 support?

Cognitive State Tracker - 通用认知状态追踪系统 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Cognitive State Tracker - 通用认知状态追踪系统?

It is built and maintained by justlikeeth (@justlikeeth); the current version is v2.0.0.

💬 Comments