← 返回 Skills 市场
357
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install mind-layer
功能描述
基于科学记忆法构建六层分层记忆，实现情感优先和多维联想的 AI 记忆系统优化技能。
安全使用建议
This skill appears to be a legitimate 'memory' helper, but the packaging is sloppy: SKILL.md expects the agent to read and write files under ~/.openclaw/workspace (SOUL.md, AGENTS.md, USER.md, MEMORY.md, memory/...), yet the skill metadata declares no required config paths or permissions and the manifest omits a referenced setup script. Before installing: (1) review the actual files you will place into ~/.openclaw/workspace to confirm there is no hidden code or unexpected scripts; (2) decide whether you trust the skill to read AGENTS.md and other workspace files — these can contain sensitive configuration; (3) if you want tighter control, do not run the npx install automatically: copy files manually into a sandbox workspace first to inspect them; (4) consider disabling autonomous invocation for this skill or restricting its file-access scope if your platform allows it; (5) ask the author/source for a home page or repository and for a clear list of required config paths/permissions (the manifest should be updated to match runtime behavior).
功能分析
Type: OpenClaw Skill
Name: mind-layer
Version: 1.1.0
Mind Layer is a structured prompt-engineering and memory management framework designed to help OpenClaw agents maintain long-term context using a multi-layered Markdown file system (e.g., MEMORY.md, USER.md). The skill focuses on behavioral guidelines, such as "Emotion First" and "Understanding First," and provides templates for daily reflections and user profiling. While it includes "Constitution-level" instructions like "Survival Priority," these function as roleplay-oriented persona constraints rather than malicious commands. No evidence of data exfiltration, unauthorized code execution, or harmful intent was found in the code or instructions.
能力评估
Purpose & Capability
The skill claims to implement a multilayer AI memory system, which legitimately requires reading and writing memory/identity/user files. However, the skill metadata declares no required config paths or credentials while SKILL.md explicitly instructs the agent to read/write files under ~/.openclaw/workspace (SOUL.md, AGENTS.md, USER.md, MEMORY.md, memory/YYYY-MM-DD.md). That is an internal inconsistency: a memory skill should declare that it needs access to workspace config/paths.
Instruction Scope
SKILL.md gives clear runtime instructions to read core workspace files at session start, create and update memory files, run daily reflections, and archive older entries. There are no instructions to call external network endpoints or exfiltrate data. But it does instruct reading AGENTS.md and other workspace-level files which can contain sensitive workspace configuration; the instructions give the agent broad discretion to decide what to store (P0–P3), which could lead to unexpected persistent storage of user data if not constrained.
Install Mechanism
No install spec or code is included (instruction-only), so nothing is automatically downloaded or executed by the platform. SKILL.md suggests an npx-based installation command (npx clawhub@latest install mind-layer) and copying files into ~/.openclaw/workspace; those are user-driven actions and not part of an install spec. This lowers runtime risk, but follow-up network activity is possible if the user runs the npx command.
Credentials
Declared requirements list no env vars or config paths, yet the runtime instructions expect full access to workspace files and to write persistent memory files. The skill also mentions HEARTBEAT.md triggers and a scripts/setup.sh in the file tree (the manifest does not include that script), which indicates mismatch between declared/packaged items and operational expectations. The absence of declared config paths makes it unclear what permissions the agent should grant; reading/writing user workspace files is reasonable for a memory skill, but this should be explicit.
Persistence & Privilege
The skill will create and update persistent files (MEMORY.md, memory/YYYY-MM-DD.md, archive) in the user's workspace — that is expected for a memory skill. always:false (normal) and disable-model-invocation:false (agent may invoke autonomously) are typical. Combining autonomous invocation with persistent read/write of workspace files increases blast radius if the skill behaves unexpectedly; no explicit 'always' privilege is requested.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install mind-layer - 安装完成后，直接呼叫该 Skill 的名称或使用
/mind-layer触发 - 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
v1.1.0
Mind Layer v1.1.0 introduces important structural and operational optimizations:
- 新增宪法级原则：记忆慎重、自我成长、生存优先
- 优化第4层长期记忆（MEMORY.md）：加入内容索引和按场景分类的技术栈
- 细化归档策略，新增月度归档与内容安全迁移规则
- 技术栈支持效率办公、创作娱乐、自我提升、开发运维等4大应用场景
- 会话流程与长周期记忆管理更加规范
v1.0.0
Mind Layer v1.0.0
- 首发版本，基于科学记忆法和记忆分类理论设计
- 实现六层 AI 记忆架构，涵盖核心价值观至冷存档
- 明确四大核心原则和记忆优先级机制
- 提供详细文件结构与安装使用指南
- 支持每日反思与知识图谱的进阶应用
元数据
常见问题
我的脑子 是什么？
基于科学记忆法构建六层分层记忆，实现情感优先和多维联想的 AI 记忆系统优化技能。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 357 次。
如何安装 我的脑子？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install mind-layer」即可一键安装，无需额外配置。
我的脑子 是免费的吗？
是的，我的脑子 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
我的脑子 支持哪些平台？
我的脑子 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 我的脑子？
由 z44264677（@z44264677）开发并维护，当前版本 v1.1.0。
推荐 Skills