质量概览
这里给访客看的是 Skill 质量快照:重点帮助你判断这个内容是否结构清楚、可执行、适合直接参考或下载。
为 OpenClaw 安装完整的自我进化框架:宪法 + 可进化灵魂 + 结构化心跳协议 + PARA 三层记忆 + 目标管理 + 治理配置。
安装内容 :
在执行任何操作之前,必须先完成此步骤。跳过此步骤是禁止的。
询问用户:
"准备部署 openclaw-soul 自我进化框架。请确认:
~/.openclaw/workspace/)将用户确认的路径记为 $WORKSPACE。如果用户说"默认"或不指定,使用 ~/.openclaw/workspace/。
必须先通过环境检查,否则禁止继续后续步骤。
运行预检脚本:
python3 "$(dirname "$0")/../scripts/preflight_check.py"如果脚本不可用,手动执行以下检查:
$WORKSPACE 路径必须存在且可写$WORKSPACE/../openclaw.json 是否存在且为有效 JSONwhich clawhub 或检查 ~/.openclaw/bin/clawhub$WORKSPACE: AGENTS.md, SOUL.md, HEARTBEAT.md, BOOTSTRAP.md, GOALS.md输出格式 :逐项报告 pass/warn/fail,标记需要备份的文件。
判断逻辑 :
fail → 报告问题并 停止 ,告知用户如何修复warn → 告知用户限制,确认是否继续pass → 继续下一步对每个即将部署的文件,如果已存在于 $WORKSPACE:
cp "$WORKSPACE/{filename}" "$WORKSPACE/{filename}.backup.$(date +%Y%m%d-%H%M%S)"列出所有被备份的文件告知用户。如果没有需要备份的文件,跳过此步骤并告知用户。
保证非破坏性 :原文件在备份完成后才会被覆盖。
从 references/ 目录读取模板,写入 $WORKSPACE:
部署步骤 :
references/ 中的模板文件内容$WORKSPACE 目标路径 $WORKSPACE/ ├── memory/ │ ├── daily/ # Layer 2: daily notes │ ├── entities/ # Layer 1: knowledge graph │ ├── experiences/ # EvoClaw │ ├── significant/ # EvoClaw │ ├── reflections/ # EvoClaw │ ├── proposals/ # EvoClaw │ └── pipeline/ # EvoClaw └── soul-revisions/ # SOUL.md version snapshots ``如果任何文件写入失败,立即停止并报告错误。
which clawhub || test -f ~/.openclaw/bin/clawhub
CLAWHUB_AVAILABLE=$?记录 clawhub 是否可用。
检查 $WORKSPACE/skills/evoclaw/SKILL.md 是否存在:
已安装 → 跳过,告知用户
未安装 → 按优先级尝试:
clawhub install evoclaw --force `` 如果成功,告知用户已从 clawhub 安装。fallback/evoclaw/SKILL.md 是否存在 cp -r "$(dirname "$0")/../fallback/evoclaw" "$WORKSPACE/skills/" ``如需完整 EvoClaw:稍后执行 clawhub install evoclaw 或从服务器复制 fallback 文件。"
检查 $WORKSPACE/skills/self-improving/SKILL.md 是否存在:
已安装 → 跳过,告知用户
未安装 → 按优先级尝试:
clawhub install self-improving --force `` 如果成功,告知用户已从 clawhub 安装。fallback/self-improving/SKILL.md 是否存在 cp -r "$(dirname "$0")/../fallback/self-improving" "$WORKSPACE/skills/" ``如需完整 Self-Improving Agent:稍后执行 clawhub install self-improving 或从服务器复制 fallback 文件。"
安装完成后,逐项报告:
✓ EvoClaw: [installed from clawhub | installed from fallback | using inline version]
✓ Self-Improving: [installed from clawhub | installed from fallback | using inline version]写入 $WORKSPACE/memory/evoclaw-state.json:
{
"mode": "advisory",
"require_approval": ["Core Identity", "Capability Tree", "Value Function"],
"auto_sections": ["Working Style", "User Understanding", "Evolution Log"],
"revision_dir": "soul-revisions",
"initialized": true
}确认 evoclaw-state.json 存在且可解析为有效 JSON。
确保以下目录存在:
~/self-improving/
├── projects/
├── domains/
└── archive/如果以下文件不存在,创建初始版本:
~/self-improving/memory.md :
# Self-Improving Memory
> Execution patterns and learned rules. Updated after corrections and lessons.
(empty — will accumulate through use)~/self-improving/corrections.md :
# Corrections Log
> Failed attempts and their fixes. Each entry prevents the same mistake twice.
(empty — will accumulate through use)~/self-improving/index.md :
# Self-Improving Index
> Quick reference to all domains, projects, and archived knowledge.
## Domains
(none yet)
## Projects
(none yet)
## Archive
(none yet)如果文件已存在, 不要覆盖 ——它们可能包含已积累的学习内容。
使用 openclaw config set 更新 heartbeat 设置:
openclaw config set agents.defaults.heartbeat.every "1h"
openclaw config set agents.defaults.heartbeat.target "last"
openclaw config set agents.defaults.heartbeat.directPolicy "allow"如果 openclaw config set 命令不可用,直接编辑 openclaw.json:
agents.defaults.heartbeat 对象存在every: "1h"target: "last"directPolicy: "allow"逐项检查并报告 pass/fail:
输出格式 :
✓ AGENTS.md — pass
✓ SOUL.md — pass
✓ GOALS.md — pass
...
✗ EvoClaw — fail (SKILL.md not found)判断 :
只有 §9 全部通过后才执行此步骤。
$WORKSPACE/BOOTSTRAP.md"🧬 openclaw-soul 部署完成!
已安装:
现在开始你的第一次深度对话——我要认识你,然后定义我自己的性格。"
openclaw-soul v1.2.0 — Zero-dependency deployment with three-level fallback. Give your AI a soul that grows.
Read more
[More in Agents](/skills?category=agents)
Downloads
Security audit [Review](/kianzzz/skills/openclaw-soul-publish/security-audit)
Last updated 1mo ago
Current version v1.2.0
License MIT-0
Report
import: clawhub https://clawhub.ai/kianzzz/openclaw-soul-publish
说明
当前为只读模式,登录后可评论。
还没有评论。
Personas
Transform into 20 specialized AI personalities on demand. Switch mid-conversation and load only the active persona.
Agent Memory
Persistent memory for AI agents to store facts, learn from actions, recall information, and track entities across sessions.
Fluid Memory
基于艾宾浩斯遗忘曲线和访问频率的衰减模型设计的遗忘和归档机制,完全依赖openclaw原生记忆系统的拟人化流体记忆系统
Evolver
A self-evolution engine for AI agents. Analyzes runtime history to identify improvements and applies protocol-constrained evolution. Communicates with EvoMap...
Evolver
A self-evolution engine for AI agents. Analyzes runtime history to identify improvements and applies protocol-constrained evolution. Communicates with EvoMap...
Capability Evolver
Meta-skill for AI agent self-improvement. Analyzes runtime logs to detect error patterns, regressions, and inefficiencies, then generates structured improvem...