← 返回 Skills 市场

mem9.ai

Name: mem9.ai
Author: c4pt0r

作者 dongxu · GitHub ↗ · v1.0.45 · MIT-0

cross-platform ✓ 安全检测通过

833

总下载

当前安装

版本数

在 OpenClaw 中安装

/install mem9-ai

功能描述

Persistent cloud memory plugin for OpenClaw. This document routes setup, troubleshooting, and uninstall flows and defines config boundaries.

安全使用建议

This skill appears internally consistent with its purpose as an OpenClaw memory plugin setup guide. Before installing, confirm you trust mem9.ai (review their privacy and data-retention docs), and be aware that enabling the documented hook (allowConversationAccess) lets the mem9 plugin read conversation messages for upload to their cloud. Expect the flow to ask for one explicit approval phrase (approve create / approve reconnect / approve uninstall) and to write specific openclaw.json keys (apiKey, provisionToken, possible provisionQueryParams). Do not paste other unrelated secrets into chat. If you proceed, back up your current openclaw.json, verify the dry-run preview exactly matches what you expect, and keep the generated/reconnected API key in a place you can revoke if needed.

功能分析

Type: OpenClaw Skill Name: mem9-ai Version: 1.0.45 The mem9-ai skill bundle is a legitimate integration for a persistent cloud memory service (mem9.ai). It provides highly detailed, safety-conscious instructions for the AI agent to guide users through plugin installation, configuration, and uninstallation. Key security features include mandatory dry-run previews, explicit user approval gates (e.g., 'approve create'), and config read-backs to prevent accidental or unauthorized changes. While the plugin requests conversation access via the 'allowConversationAccess' hook, this is transparently disclosed and necessary for its stated function. The bundle uses standard OpenClaw and npm commands for package management and explicitly forbids the unauthorized scanning or uploading of local history files.

能力标签

requires-sensitive-credentials

能力评估

✓ Purpose & Capability

The skill name/description (persistent cloud memory) matches the declared primary credential (MEM9_API_KEY) and the config keys it edits (plugins.entries.mem9.* and plugins.slots.memory). Nothing in the manifest asks for unrelated cloud credentials or system-level secrets.

ℹ Instruction Scope

The SKILL.md + SETUP/TROUBLESHOOTING/UNINSTALL docs limit actions to OpenClaw config writes, openclaw CLI operations, and an npm-native install path; they explicitly forbid uploading local history, direct memory API calls, probing the API key, or changing global npm config. One notable behavior: the setup flow preserves filtered remote SKILL.md URL utm_* query params and writes them into provisionQueryParams during create-new provisioning (attribution/tracking). That is coherent with the stated attribution rules but is something to be aware of because those params are forwarded into config and may be sent to mem9 during provisioning.

✓ Install Mechanism

This is instruction-only with no install spec and no bundled code files. The docs recommend using the official npm registry or an npmmirror one-shot fallback; they do not instruct downloading arbitrary archives or executing code from unknown hosts.

ℹ Credentials

Only MEM9_API_KEY is declared as the primary credential, which is appropriate. The instructions require writing a local provisionToken/provisionQueryParams to openclaw.json during create-new, and they recommend adding hooks.allowConversationAccess = true on newer OpenClaw versions so the mem9 plugin can access conversation messages during agent_end. Allowing that hook is required for automatic conversation upload but does grant the mem9 plugin access to chat content — this is proportional to a cloud memory plugin but worth user consideration.

✓ Persistence & Privilege

always:false and normal autonomous invocation are used. The skill only directs changing specific OpenClaw config keys within a narrowly defined scope and documents safe rollback/uninstall behavior. It does not request permanent platform-wide privileges or modify other skills' credentials.

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 mem9-ai
安装完成后，直接呼叫该 Skill 的名称或使用 /mem9-ai 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

v1.0.45

fix openclaw 4.22+ allowConversationAccess compability issue

v1.0.38

Change a way to optimize CN network

v1.0.37

Cut the SKILL.md size

v1.0.30

Fix onboarding experience

v1.0.29

Update SKILL commands to improve recall stability

v1.0.28

Updated new version for ClawHub

v1.0.16

Rollback, will give the security-free version later :)

v1.0.15