← 返回 Skills 市场
595
总下载
0
收藏
4
当前安装
30
版本数
在 OpenClaw 中安装
/install memory-palace
功能描述
持久化记忆管理。Use when: 用户告诉你个人信息/偏好/习惯、需要记住项目状态/技术决策、完成任务后有可复用经验、用户说"记住""别忘了""下次注意"、需要回忆之前的对话内容。支持语义搜索和时间推理。
安全使用建议
This skill appears to be what it claims: a local persistent memory manager with optional semantic search. Before installing, be aware of these practical implications: (1) it stores memories as Markdown files under a workspace (default ~/.openclaw/workspace/memory/palace) — do not store highly sensitive secrets there unless you encrypt them; (2) optional vector search will auto-download an embedding model (~100MB) from HuggingFace (or a mirror if HF_ENDPOINT is set) and requires Python + sentence-transformers; this involves network traffic and disk use; (3) npm install will run a postinstall script (scripts/check-vector-deps.cjs) — review package.json and that script if you want to inspect install-time behavior; (4) the CLI reads OPENCLAW_WORKSPACE and HOME to pick paths and supports a --db-path override; consider running the skill in an isolated workspace or container if you want to limit its filesystem scope. If you want, review scripts/check-vector-deps.cjs and scripts/vector-service.py to confirm model download behavior and any network endpoints before installing.
功能分析
Type: OpenClaw Skill
Name: memory-palace
Version: 1.8.5
The memory-palace skill bundle is a well-architected persistent memory system for OpenClaw agents. It provides CRUD operations for memories, semantic search via a local Python-based vector service (scripts/vector-service.py), and LLM-enhanced features like summarization and experience extraction. Security-wise, the code demonstrates good practices, such as using regex escaping in src/manager.ts to prevent injection attacks during text searches. The installation script (scripts/check-vector-deps.cjs) is transparent and prompts the user before installing Python dependencies. No evidence of data exfiltration, unauthorized remote execution, or malicious prompt injection was found; all behaviors are strictly aligned with the stated purpose of memory management.
能力评估
Purpose & Capability
Name/description (persistent memory, semantic search, time reasoning) match the provided code, CLI, and files. Required binary is 'node' and the package provides a 'memory-palace' CLI — this is proportionate to the stated purpose. No unexpected credentials or unrelated service access are requested.
Instruction Scope
SKILL.md and CLI instruct the agent to store/read Markdown files under a workspace (default: ~/.openclaw/workspace/memory/palace) and optionally start a Python-based vector service that will auto-download a ~100MB embedding model (BAAI/bge-small-zh-v1.5). These operations write to disk and perform network downloads for models; they are reasonable for semantic search but constitute observable side effects that users should accept before installing.
Install Mechanism
Install spec is an npm package 'memory-palace' (provides the 'memory-palace' binary) — a common/traceable mechanism. package.json includes a postinstall script (node scripts/check-vector-deps.cjs) which will execute during npm install; this is expected for packages that verify optional dependencies but increases install-time execution risk compared with an instruction-only skill. No download-from-arbitrary-URL installer was observed.
Credentials
The skill does not require secrets or unrelated environment variables. It does read OPENCLAW_WORKSPACE or HOME to determine storage location and mentions optional env vars for model mirrors (HF_ENDPOINT) and some feature toggles in docs; these are reasonable and proportional to a local memory manager.
Persistence & Privilege
Skill is not 'always:true' and does not request persistent platform-level privileges. It creates and manages its own workspace files under the user's workspace/home directory, which is expected for a file-backed memory skill.
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 memory-palace - 安装完成后，直接呼叫该 Skill 的名称或使用
/memory-palace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
v1.8.5
Bug 2: CLI 参数解析修复; Bug 5: vector-search ESM __dirname 修复 + Python 模型路径修复
v1.8.4
- Updated documentation for clearer installation steps, specifically moving semantic search enhancement details earlier and clarifying Python usage.
- Removed a redundant "安装 Skill" section from the documentation.
- No changes to functionality; this release is documentation-focused for better user guidance.
v1.8.3
- Clarified installation instructions: now recommends installing the Skill via `npx clawhub install memory-palace`.
- Updated the location for downloaded embedding models to `~/.openclaw/models/embedding/`.
- Removed redundant CLI installation section to avoid confusion.
- No changes to core functionality or parameters.
v1.8.1
- Added installation instructions for CLI and semantic search dependencies to the documentation.
- Clarified optional Python and model requirements for semantic search, including auto-download details.
- Noted automatic fallback to text search if Python is unavailable.
- No functional or API changes; documentation only.
v1.8.0
**Summary:** Major upgrade with architecture refactor, improved documentation, and new cognitive background modules.
- Major documentation overhaul: added usage examples, API reference, and architecture docs; streamlined SKILL.md for clarity and practical usage.
- New cognitive modules: added duplicate detection, importance evaluator, tag inference, and time pattern extraction to enhance memory processing.
- Simplified experience verification: experience validity now uses an intuitive effectiveness score mechanism.
- File structure reorganized: removed obsolete docs, tests, and scheduler code; added TypeScript migration for CLI.
- Improved metadata and install process for better OpenClaw integration.
- Various core files updated for stability, extensibility, and enhanced background reasoning.
v1.6.5
**This update greatly expands the documentation with full tool parameter details and examples.**
- SKILL.md overhauled to document all tools, parameters, and usage examples
- Added explicit descriptions for each memory-palace API, including required/optional parameters
- Included documentation for new tools: memory_palace_get_frequently_accessed, memory_palace_record_access, memory_palace_check_model_status
- Existing feature and usage notes carried forward and clarified
- No code or functional changes—documentation improvements only
v1.6.4
- Documentation updated to reflect latest features under v1.6.4.
- Changelog and feature highlights moved from v1.6.2 to v1.6.4 in documentation.
- No changes to features or functionality; this version is a documentation update only.
v1.6.3
- Bumped version to 1.6.3.
- Updated documentation: "新增功能" now references v1.6.2 as baseline and removes previous 1.3.9 features section.
- No new user-facing features or breaking changes introduced.
v1.6.2
v1.6.2: Add ESLint, CI/CD, auto-compression scheduler, access tracking, and improved vector model installation
v1.6.1
**Version 1.6.1 Highlights:**
Improves experience scoring and introduces memory linking/relationship features.
- Added support for linking memories and retrieving related memories via `memory_palace_link` and `memory_palace_get_related`.
- Enhanced experience management: now includes `effectivenessScore` with detailed scoring rules for usage and verification.
- Updated documentation with guidance on installing vector search dependencies for improved semantic search.
- Removed deprecated or outdated files (e.g., `_meta.json`).
- General documentation and parameter explanation improvements for clarity.
v1.6.0
发布到 memory-palace slug，统一命名规范
v1.5.0
艾宾浩斯遗忘曲线实现，支持环境变量配置。新增 getDecayStats/getForgottenMemories 方法。SKILL.md 优化为场景指南。
v1.4.3
Add .clawhubignore to exclude unnecessary files. Reduce package size.
v1.4.2
LLM tools: extract_experience, expand_concepts, compress
v1.4.1
修复：包含完整源代码
v1.4.0-beta.3
- Major documentation overhaul: SKILL.md was fully rewritten, providing clearer tool descriptions, parameters, and usage examples.
- Outdated, duplicated, and auxiliary documentation files were removed for better clarity and maintenance.
- Tool list and parameter formats were streamlined and clarified, with a new section for experience statistics tools.
- Version bumped to 1.4.0, reflecting significant changes in documentation and skill metadata.
v1.4.0
- Reverted version to 1.3.7 and restored previous SKILL.md structure.
- Removed beta indicators from documentation and reintroduced stable references section.
- Added new LLM-related tools (e.g., memory_palace_extract_experience, memory_palace_expand_concepts_llm) and auxiliary tools to the tool list.
- Restructured documentation to refer to detailed tool parameters and usage examples in separate reference files.
- Removed RELEASE_BETA.md and related beta documentation.
- Added test-llm-features.cjs for new LLM feature testing.
v1.4.0-beta.2
Version 1.4.0-beta.2
- Major refactor of documentation, including a new RELEASE_BETA.md and streamlined SKILL.md for clarity and updated tool details.
- Expanded and clarified tool APIs: added fields, improved parameter descriptions, and reviewed tool lists.
- Introduced or improved experience statistics command (memory_palace_experience_stats).
- Updated tool names and removed deprecated/inactive LLM tools (e.g., removed concept expand/compress).
- Package and script updates to support new structure and enhanced experience/statistics features.
- General improvements for usability, maintainability, and onboarding instructions (vector model installation emphasized).
v1.3.8
memory-palace v1.3.8
- Added detailed tool reference (references/tools.md) and example (references/examples.md) documentation.
- Modularized documentation: SKILL.md now links to new references for tools and examples.
- Added initial CLI test (tests/cli.test.js) to improve test coverage.
- Updated package metadata: license field and minor version adjustment.
- General doc cleanup and structure improvements.
v1.3.7
fix CLI update and record_experience parameter passing bugs
元数据
常见问题
Memory Palace 是什么？
持久化记忆管理。Use when: 用户告诉你个人信息/偏好/习惯、需要记住项目状态/技术决策、完成任务后有可复用经验、用户说"记住""别忘了""下次注意"、需要回忆之前的对话内容。支持语义搜索和时间推理。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 595 次。
如何安装 Memory Palace？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install memory-palace」即可一键安装，无需额外配置。
Memory Palace 是免费的吗？
是的，Memory Palace 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Memory Palace 支持哪些平台？
Memory Palace 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 Memory Palace？
由 蓝宙（@lanzhou3）开发并维护，当前版本 v1.8.5。
推荐 Skills