马斯克精英记忆

终极AI智能体记忆系统。WAL协议 + 向量搜索 + git-notes + 云备份。6层记忆架构：热RAM(SESSION-STATE)、温存储(LanceDB向量)、冷存储(Git-Notes知识图)、精选档案(MEMORY.md)、云备份(SuperMemory)、自动提取(Mem0)。

This skill appears to be a legitimate local memory helper that writes persistent files and offers optional cloud integrations, but there are several mismatches you should consider before installing: - Secrets & cloud services: SKILL.md and README reference SUPERMEMORY_API_KEY, MEM0_API_KEY, and an OpenAI key, but the registry metadata declares no required env vars. If you enable SuperMemory, Mem0, or OpenAI integrations you will supply API keys that could allow memory data to be sent to external services. Only enable cloud backups if you trust those services and are comfortable that your saved memories may be transmitted. - Persistent storage: The skill enforces a write-ahead rule (write to SESSION-STATE.md before responding). That means conversational context (including sensitive details) may be written to files in your workspace. Review and control what you store there; do not let the agent write secrets (passwords, keys, PII) into these files. - Destructive maintenance command: The README/ SKILL.md show a command to clear vectors: rm -rf ~/.openclaw/memory/lancedb. That will delete local data; back up before running, and avoid running such commands unless you understand the consequences. - Metadata inconsistencies: The skill's registry metadata lists no env requirements and an unknown source/homepage, while package.json points to a GitHub repo and _meta.json shows a different ownerId. Treat source provenance as uncertain; prefer packages with a verifiable repository and author identity. - Safe testing recommendations: Run npx elite-longterm-memory init in an isolated test directory (not your real workspace), inspect the created files, and verify behavior. If you plan to enable cloud integrations, do so in a controlled environment and review what data is exported (use minimal/test accounts/keys). Consider auditing any optional library (mem0ai) before installing and do not add secrets into MEMORY.md or SESSION-STATE.md. If you want, I can list the exact lines where undeclared env vars and the rm -rf appear, or produce a checklist to safely evaluate and sandbox this skill before enabling cloud backups.

Type: OpenClaw Skill Name: maske-elite-memory Version: 1.0.0 The skill bundle provides a comprehensive multi-layered memory management system for AI agents, utilizing local Markdown files (SESSION-STATE.md, MEMORY.md), vector databases (LanceDB), and external APIs (Mem0, SuperMemory). The included Node.js CLI tool (bin/elite-memory.js) safely initializes the workspace file structure and checks for existing memory directories. While the instructions in SKILL.md suggest using external services and performing background logging, these actions are clearly aligned with the stated purpose of persistent context management and do not exhibit signs of malicious intent or unauthorized data exfiltration.