← 返回 Skills 市场
Mem0 Memory Layer
作者
Tang Weigang
· GitHub ↗
· v0.1.0
· MIT-0
63
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install mem0-memory-layer
功能描述
Mem0 长期记忆层：为 LLM agent / chatbot 提供事实级记忆——抽取、嵌入、去重、存储 + 混合检索（语义 + BM25 + 实体加权），覆盖 17 个核心用例。自托管 Memory 与托管 MemoryClient 双形态。 Mem0 long-term memory layer for L...
安全使用建议
This skill bundles a Mem0 memory description with a large compiled blueprint (seed.yaml) that instructs the host to run Python import checks, access and write workspace files, and rely on environment values (OpenAI keys, MEM0_API_KEY, MEM0_DIR, ZVT_HOME) that are not declared in the registry. Before installing or enabling it: (1) review seed.yaml fully to confirm which commands/files it will read or write, (2) do not supply API keys or secrets unless you trust the source — run it in an isolated environment, (3) if you want to use only Mem0 docs, extract and inspect the specific mem0 parts and ignore unrelated finance/backtest preconditions (zvt), (4) explicitly set MEM0_TELEMETRY=false if you do not want telemetry/analytics, and (5) ask the publisher for a clear, minimal manifest declaring required env vars, install steps, and exact filesystem operations. The current package looks like a mixed/compiled blueprint rather than a focused memory-only skill — proceed with caution.
功能分析
Type: OpenClaw Skill
Name: mem0-memory-layer
Version: 0.1.0
The skill bundle exhibits a significant functional discrepancy: while the metadata and SKILL.md describe a 'Mem0' memory layer, the core logic in 'references/seed.yaml' is almost entirely dedicated to the 'ZVT' quant trading framework (finance-bp-131). Most notably, the skill includes default-on telemetry (BD-019, mem0-C-004) that exfiltrates MD5-hashed caller IDs and lifecycle data to a hardcoded PostHog endpoint. While this behavior is documented within the constraints, the mismatch between the stated purpose and the actual financial trading execution logic, combined with broad shell execution permissions for environment setup, presents a high risk of unintended behavior.
能力标签
能力评估
Purpose & Capability
The SKILL.md describes Mem0 (an LLM memory layer) which legitimately may need Python, a local MEM0_DIR, and optional MEM0_API_KEY. However the included seed.yaml is a compiled blueprint with id finance-bp-131 and contains domain-specific preconditions (zvt package, ZVT_HOME, finance backtest checks) that are unrelated to a generic memory layer. Registry metadata declares no required env vars or config paths despite the SKILL.md and seed.yaml referencing Python runtime, local directories (~/.mem0 or ZVT_HOME), and optional API keys — this mismatch is incoherent.
Instruction Scope
The skill’s runtime instructions (seed.yaml) require the host agent to reload seed.yaml on every decision, run precondition checks that execute Python commands (import checks, filesystem checks), verify packages before proceeding, and write into host_workspace (scripts/, skills/, .trace/). These actions go beyond a read-only documentation skill and instruct executing commands and touching local files; they also reference LATEST.yaml/LATEST.jsonl lookups and other artifacts not declared in the skill metadata. That scope creep is disproportionate and surprising for a memory-layer skill.
Install Mechanism
No install spec is provided (instruction-only), which is low risk in itself. However the execution_protocol in seed.yaml instructs the host to run install recipes and verify imports at runtime (e.g., 'Execute resources.host_adapter.install_recipes[]' and python import checks). Those runtime installs are not declared in the registry metadata, so while there is no packaged installer, the instructions effectively request package installation at runtime — this is noteworthy.
Credentials
The registry lists no required environment variables, yet SKILL.md text and seed.yaml refer to multiple environment/config items (Python 3.10+, OpenAI as default LLM/embedding provider, MEM0_DIR ~/.mem0, optional MEM0_API_KEY, MEM0_TELEMETRY, and ZVT/ZVT_HOME checks). Requesting access to local directories, possible API keys, and unrelated project-specific variables (ZVT) without declaring them is disproportionate and raises a risk of unexpected credential or filesystem access.
Persistence & Privilege
always:false and user-invocable:true (normal). The seed.yaml instructs creating/using host_workspace paths (scripts/, skills/, .trace/) and mandates reloading seed.yaml on behavioral decisions — this implies persistent presence in the workspace but not an elevated platform-wide privilege. It's not an outright privilege escalation, but it does ask to write into the agent workspace and to rely on filesystem traces which users should be aware of.
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 mem0-memory-layer - 安装完成后，直接呼叫该 Skill 的名称或使用
/mem0-memory-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
v0.1.0
Initial release: Mem0 long-term memory layer skill (52 constraints / 1 fatal). Doramagic.ai/zh/crystal/mem0-memory-layer
元数据
常见问题
Mem0 Memory Layer 是什么？
Mem0 长期记忆层：为 LLM agent / chatbot 提供事实级记忆——抽取、嵌入、去重、存储 + 混合检索（语义 + BM25 + 实体加权），覆盖 17 个核心用例。自托管 Memory 与托管 MemoryClient 双形态。 Mem0 long-term memory layer for L... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 63 次。
如何安装 Mem0 Memory Layer？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install mem0-memory-layer」即可一键安装，无需额外配置。
Mem0 Memory Layer 是免费的吗？
是的，Mem0 Memory Layer 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Mem0 Memory Layer 支持哪些平台？
Mem0 Memory Layer 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 Mem0 Memory Layer？
由 Tang Weigang（@tangweigang-jpg）开发并维护，当前版本 v0.1.0。
推荐 Skills