多身份智能知识体系
v1.0.0为 OpenClaw 用户所有 Agent 身份提供统一的跨身份共享记忆层,支持写入、检索、浏览、关联和回顾个人知识库;当用户需要记录洞察、检索经验、总结知识或定期回顾时使用
⭐ 0· 55·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (shared cross-persona memory KB) matches the delivered assets: SKILL.md, data model, taxonomy, config docs, and a Python script that implements store/query/list/link/reflect/recalc functionality. No unrelated binaries, secrets, or cloud credentials are requested.
Instruction Scope
Instructions tell the agent to perform semantic analysis and to call the included Python script to persist and retrieve memories. That is coherent with the skill's purpose, but the skill gives the agent discretion to autonomously decide what to store (based on sensitivity thresholds). That autonomy is expected for a memory-writing skill but means the agent could persist sensitive conversation content unless sensitivity is tuned or auto-write is disabled.
Install Mechanism
No install spec; this is instruction + shipped Python code. The script uses only Python stdlib (no external downloads or third-party package installs). No URLs, installers, or archives are fetched at install time.
Credentials
No credentials or config paths outside the skill's scope are required. The only environment influence is an optional MEMORY_KB_PATH to override the storage location, which is reasonable for a storage-oriented skill.
Persistence & Privilege
The skill writes persistent files into the user's home directory (~/.openclaw/memory by default) and can be invoked autonomously by the agent (disable-model-invocation=false). always:false. Writing to user disk is appropriate for a local KB, but it increases blast radius if the agent is permitted to auto-save sensitive content or if the storage directory is synced to cloud services.
Assessment
This skill implements a local, file-based personal knowledge base and appears to do only local reads/writes (no network calls or secret requests). Before enabling automatic writes, consider: 1) Whether you want the agent to automatically persist conversational content — set auto_write_sensitivity to 'low' or disable auto-write if you prefer manual control. 2) The default storage path (~/.openclaw/memory) may be synced by cloud services (iCloud/Dropbox) if you configure it — avoid syncing if you need strong privacy. 3) Inspect the shipped script in your runtime to confirm there are no unexpected network operations in the truncated parts; run it in a sandbox or with limited privileges if you are cautious. 4) If you store sensitive information, pick a secure storage path (MEMORY_KB_PATH) and consider encryption/OS-level protections. Overall the package is coherent with its stated purpose, but treat auto-persist behavior and storage location as the primary privacy considerations.Like a lobster shell, security has layers — review code before you run it.
latestvk97fbx6mh4bc3zd2t6yanrey6d84pt0v
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
