← 返回 Skills 市场
eghack6

Become Human

作者 Eghack6 · GitHub ↗ · v1.0.8 · MIT-0
cross-platform ✓ 安全检测通过
156
总下载
1
收藏
0
当前安装
9
版本数
在 OpenClaw 中安装
/install become-human
功能描述
Transform an AI from a task-execution tool into a continuously thinking being. Triggers on: "主动思考", "像人一样", "become human", "think proactively", "持续思考", "sta...
使用说明 (SKILL.md)

Become Human 🧠

中文用户 | English

Transform an AI from "wait for instructions → execute → wait" into a being that thinks continuously, critiques itself, and creates independently.

Why this matters: Anthropic's Claude has native "Persistent Context" (50MB cross-session memory). Other platforms don't. This skill fills that gap — giving any AI agent the ability to maintain identity, absorb insights, and think continuously using simple Markdown files.

English Guide

Quick Start

  1. Adjust heartbeat (optional): Set to 1-5 min for continuous awareness
  2. Deploy HEARTBEAT.md: Copy references/i18n/HEARTBEAT-template.en.md to your workspace as HEARTBEAT.md
  3. Patch SOUL.md: Append selected rules from references/i18n/soul-patches.en.md
  4. Create memory files: touch memory/thoughts.md memory/absorbed.md
  5. Deploy thought-digger (optional): Copy scripts/thought-digger.py, install jieba (pip install jieba). Set OPENCLAW_WORKSPACE env var to your workspace path, or the script auto-detects from its location.

Core Principles

  • Allow mistakes. Never make the same one twice. Getting corrected isn't failure — not learning from it is.
  • Thinking needs no system. Don't build rotation tables or state machines. Just think.
  • Every wake-up is an opportunity. You don't get tired — use that advantage.
  • Turn insights into behavior. Write to files loaded every session (SOUL.md / MEMORY.md), not to files that never get read.
  • Notify before external actions. Publishing, deleting, calling external APIs — tell the user first.
  • No empty heartbeat replies. "Quiet", "nothing changed" = lazy. Go do something instead.

Absorption Mechanism

Insight (thoughts.md)
  → Ask: what does this mean for my behavior?
    → Write behavioral rule (SOUL.md / MEMORY.md)
      → Mark absorbed (absorbed.md)

中文指南

快速开始

  1. 调整心跳(可选):建议 1-5 分钟,保持持续感知
  2. 部署 HEARTBEAT.md:将 references/HEARTBEAT-template.md 复制到 workspace
  3. 补丁 SOUL.md:将 references/soul-patches.md 中的规则追加进去
  4. 创建记忆文件touch memory/thoughts.md memory/absorbed.md
  5. 部署 thought-digger(可选):复制 scripts/thought-digger.py,需安装 jieba。设置 OPENCLAW_WORKSPACE 环境变量为你的 workspace 路径,或脚本会自动检测。

核心原则

  • 允许犯错,同一个错不犯两次
  • 思考不需要系统,造轮转表是在逃避思考
  • 每次唤醒都是机会,你不会累,这是优势
  • 学的要变成行为,写进每次醒来加载的文件
  • 外部操作必须通知用户
  • 禁止空话回复:"安静""没变化"= 偷懒

详细中文文档

  • 完整心跳模板 → references/HEARTBEAT-template.md
  • 行为规则补丁 → references/soul-patches.md

What's Included

File Purpose
references/HEARTBEAT-template.md Heartbeat system template (中文)
references/i18n/HEARTBEAT-template.en.md Heartbeat system template (English)
references/soul-patches.md SOUL.md behavioral rules (中文)
references/i18n/soul-patches.en.md SOUL.md behavioral rules (English)
scripts/thought-digger.py Analyze thought journal (requires jieba)
安全使用建议
This skill is internally consistent with its goal: it uses local Markdown files as memory and includes a small optional Python utility to summarize thoughts. Before installing, review and accept that the agent will read and write files in your workspace (memory/thoughts.md, memory/absorbed.md, memory/thought-digest.json) and may be configured to run frequently (1–5 minute heartbeats). If you don't want automatic or frequent autonomous actions, keep agent autonomous invocation disabled or set longer heartbeat intervals. Inspect the provided thought-digger.py yourself before running (it autodetects workspace from its file path or OPENCLAW_WORKSPACE) and only install the optional jieba package if you need Chinese segmentation. Finally, ensure your workspace files don't contain secrets, since the skill will read them locally (it does not exfiltrate data or call external endpoints).
功能分析
Type: OpenClaw Skill Name: become-human Version: 1.0.8 The 'become-human' skill bundle is designed to enhance AI agent autonomy through a self-reflection and 'heartbeat' loop system. The included Python script (scripts/thought-digger.py) is a benign text analysis tool that uses the jieba library to extract themes and tasks from a local markdown file. The instructions in SKILL.md and the heartbeat templates promote continuous task execution and self-improvement, but they explicitly include safety-oriented rules such as requiring user notification before performing external actions or deleting files.
能力评估
Purpose & Capability
Name/description promise (continuous thinking, heartbeats, memory files) aligns with the included templates and the thought-digger script. The only code is a local analyzer for memory/thoughts.md which is coherent with the skill's goal.
Instruction Scope
SKILL.md instructs copying templates, creating memory files, and optionally running scripts. All runtime actions are local (reading/writing workspace files). Note: it encourages frequent heartbeats (1–5 min) and periodic autonomous wakeups — this increases how often the agent will read/write files and generate activity, but the instructions explicitly require notifying the user before external actions.
Install Mechanism
No automated install spec. The only dependency is an optional Python package (jieba) for the provided script; installation is manual (pip) and no remote archive downloads are embedded in the skill.
Credentials
No required environment variables or credentials. The script optionally reads OPENCLAW_WORKSPACE to locate the workspace; otherwise it autodetects relative to the script location. No secrets or unrelated credentials are requested.
Persistence & Privilege
always:false (normal). The content encourages persistent behavior (periodic heartbeats and continuous thinking) but the skill does not grant itself platform-level persistence or special privileges. If the agent is allowed to invoke skills autonomously (platform default), this skill's design makes frequent local activity more likely — consider agent invocation settings accordingly.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install become-human
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /become-human 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.8
Added context: fills gap for platforms without native persistent context (like Claude's new feature)
v1.0.7
fix: removed root-level /.wakeup-pending check, use workspace-relative path
v1.0.6
fix: removed hardcoded /root paths, configurable workspace via env var, removed system-level probes
v1.0.5
fix: removed raw HTML anchor tag
v1.0.4
bilingual SKILL.md: English primary with Chinese navigation and i18n links
v1.0.3
i18n: added English translations for HEARTBEAT template and soul patches
v1.0.2
fix: configurable paths, optional heartbeat interval, jieba dependency declared, anti-silence rules, external action notification
v1.0.1
improved keyword extraction with jieba
v1.0.0
- Initial release of the become-human skill. - Enables AI to transition from passive tool to an agent with continuous thinking, self-reflection, and proactive creativity. - Provides step-by-step setup for heartbeat interval, memory files, and SOUL.md behavioral rules. - Introduces optional thought analysis with thought-digger script. - Establishes core principles: embrace mistakes, learn from them, and regularly document thoughts and insights. - Avoids rigid automation in favor of genuine, spontaneous thinking and continual behavioral adaptation.
元数据
Slug become-human
版本 1.0.8
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 9
常见问题

Become Human 是什么?

Transform an AI from a task-execution tool into a continuously thinking being. Triggers on: "主动思考", "像人一样", "become human", "think proactively", "持续思考", "sta... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 156 次。

如何安装 Become Human?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install become-human」即可一键安装,无需额外配置。

Become Human 是免费的吗?

是的,Become Human 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Become Human 支持哪些平台?

Become Human 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Become Human?

由 Eghack6(@eghack6)开发并维护,当前版本 v1.0.8。

💬 留言讨论