← 返回 Skills 市场
Neural Memory CN
作者
Sunlight-Bulling
· GitHub ↗
· v1.0.1
· MIT-0
254
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install neural-memory-cn
功能描述
神经网络启发的记忆系统，支持激活扩散和联想检索。安装后即可使用本地模式，配置 LLM 后可启用智能意图分析。/ Neural network-inspired memory with activation spreading. Works out-of-box in local mode; configure L...
安全使用建议
This skill appears to implement a local neural-memory system and will read and persist memory files under a ~/.openclaw/neural-memory path. Before installing: (1) Inspect where the skill will be installed (the SKILL.md suggests ~/.openclaw/skills/... while the setup writes to ~/.openclaw/neural-memory — confirm expected paths). (2) Be cautious about supplying any LLM API keys; the code looks for multiple env vars (NEURAL_MEMORY_LLM_*, OPENROUTER_API_KEY) and may use an available OpenAI/openrouter key unexpectedly. (3) The adapter integrates with the platform's existing memory/search and maps neurons to user-related files (UserProfile.md, Preferences.md). If you store sensitive profile data in platform memory, expect this skill to surface or index it unless you change protection settings. (4) If you want to proceed, review the bundled Python files locally (especially the adapter and intent layer) and run the setup in a sandbox or test environment first. If anything is unclear, ask the skill author to clarify path and env-var behavior and how integrations with openclaw.memory are gated.
功能分析
Type: OpenClaw Skill
Name: neural-memory-cn
Version: 1.0.1
The bundle implements a sophisticated neural-network-inspired memory system for OpenClaw agents, featuring activation spreading and semantic retrieval. The code is well-structured and its behavior aligns perfectly with the stated purpose of enhancing agent memory through neuron/synapse simulation. It handles sensitive LLM API keys via standard environment variables or local configuration files (config.yaml) and includes a 'protection' mechanism to prevent the accidental deletion of core user identity and preference data. No evidence of data exfiltration, malicious execution, or prompt injection attacks was found; all network activity is directed toward user-configured LLM providers (e.g., OpenRouter or OpenAI) for legitimate intent analysis.
能力评估
Purpose & Capability
The code and README implement a local neural-memory system (neurons, synapses, activation spreading, storage). Required binaries (python) and the optional LLM integration align with the declared purpose. The presence of an adapter that merges results with OpenClaw's memory search is consistent with a memory-enhancement skill.
Instruction Scope
Runtime instructions and code operate on local storage under ~/.openclaw/neural-memory and expose APIs like think(), learn_and_think(), save(). However the ThinkingAdapter imports and calls openclaw.memory.memory_search and maps neurons to files such as UserProfile.md and Preferences.md — meaning the skill will access and reuse other system memory/search results. This is plausible for a memory-augmentation skill but increases the scope of data the skill can read and return.
Install Mechanism
There is no remote download/install URL; this is an instruction-and-code skill with Python scripts bundled. No extract-from-URL or installer that pulls arbitrary binaries was included. Installation appears to be via the platform's skill installer (npx clawhub) which places these files locally.
Credentials
LLM integration is optional and requires API keys, which is reasonable. However the skill references multiple environment variable conventions (SKILL.md suggests NEURAL_MEMORY_LLM_API_KEY/NEURAL_MEMORY_LLM_BASE_URL/NEURAL_MEMORY_LLM_MODEL; intent_layer._get_openrouter_key looks for OPENROUTER_API_KEY; some code tries to use the openai client if present). This multiplicity could cause the skill to pick up an existing LLM credential unexpectedly. No other unrelated credentials are requested.
Persistence & Privilege
The skill does not request 'always: true' and follows normal autonomy defaults. Its config templates enable integration options (e.g., memory_search_enhancement, create_thinking_endpoint, auto_link_knowledge) and the adapter constructs an integration that will call into platform memory — this gives it broader local data access but does not itself elevate system privileges. Review how the platform wires adapters/endpoints before enabling.
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 neural-memory-cn - 安装完成后，直接呼叫该 Skill 的名称或使用
/neural-memory-cn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
v1.0.1
- Added detailed setup and usage instructions in Chinese and English to SKILL.md.
- Explained all usage modes, including local/out-of-box usage and optional LLM integration.
- Provided configuration methods for LLM (environment variables, setup script, config file).
- Documented core concepts, file structure, API, and troubleshooting in an organized format.
v1.0.0
- Initial release of neural-memory: a brain-inspired memory system for smart associative query and knowledge retrieval.
- Supports activation spreading, intent understanding, and on-demand knowledge loading.
- Multiple usage modes: smart (intent-driven), exact (name lookup), and associative (hybrid).
- Flexible configuration (YAML) for memory flow, LLM integration, and storage settings.
- Data stored modularly with neuron/synapse separation and knowledge domain definitions.
- Protects designated neurons from deletion for safety.
元数据
常见问题
Neural Memory CN 是什么？
神经网络启发的记忆系统，支持激活扩散和联想检索。安装后即可使用本地模式，配置 LLM 后可启用智能意图分析。/ Neural network-inspired memory with activation spreading. Works out-of-box in local mode; configure L... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 254 次。
如何安装 Neural Memory CN？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install neural-memory-cn」即可一键安装，无需额外配置。
Neural Memory CN 是免费的吗？
是的，Neural Memory CN 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Neural Memory CN 支持哪些平台？
Neural Memory CN 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 Neural Memory CN？
由 Sunlight-Bulling（@sunlight-bulling）开发并维护，当前版本 v1.0.1。
推荐 Skills