← 返回 Skills 市场
michaelfeng

Clawbrain Boost

作者 michaelfeng · GitHub ↗ · v1.6.0 · MIT-0
cross-platform ⚠ suspicious
138
总下载
0
收藏
0
当前安装
4
版本数
在 OpenClaw 中安装
/install clawbrain-boost
功能描述
ClawBrain Boost v1.6 — 一键让 OpenClaw 更准更稳。记忆系统 + 数据保真 + 自动容错 + 写作助手。越用越懂你。
使用说明 (SKILL.md)

ClawBrain Boost

一键让你的 OpenClaw 更准、更稳、越用越懂你。

安装后你得到什么

四档性能

  • Flash(0.5 Credits)— 简单任务秒回,省钱
  • Pro(1 Credit)— 均衡性能,适合日常任务
  • Max(3 Credits)— 深度推理,返回完整思考过程
  • Auto(推荐)— 自动判断复杂度,调整推理深度

记忆系统

  • 每次对话中的重要信息自动提取为结构化实体和关系
  • 支持 6 种实体类型:人物、项目、工具、决策、偏好、问题
  • 中文 N-gram 智能分词,精准检索历史记忆
  • 每日自动整合(DreamTask 凌晨 3:00 自动运行),无需手动维护
  • 3D 可视化知识图谱,在控制台查看
  • 记忆来源标注 — 每条记忆附带来源对话引用,可追溯
  • 身份信息自动更新 — 检测到用户身份变化时自动同步
  • 长对话不丢上下文 — 超长对话压缩时自动从记忆恢复关键信息

数据保真

  • 你给的数字/日期/人名不会被 AI 篡改
  • 自动提取锁定,生成后逐一校验
  • 改了就打回重写

响应更好

  • 简洁直接,不寒暄不废话
  • 模糊指令先确认理解再动手
  • 高风险操作先征得同意
  • 出错自动修复,切换策略恢复
  • 主动思考框架:执行前自问"不知道什么/可能出问题/怎么验证"

垂直场景加持

自动识别任务领域,注入专业规则:

  • 支付场景 → 提醒幂等性、签名验证
  • 运维场景 → 提醒先备份再操作
  • 代码场景 → 检查安全漏洞、边界条件
  • 数据场景 → 提醒数据完整性、脱敏

输出质量保障

  • 独立四维评分:准确性、完整性、逻辑性、格式
  • 评分低于 70 分自动重试
  • 99.9% 可用性

安装

clawhub install clawbrain-boost

手动配置

编辑 ~/.openclaw/openclaw.json

{
  "models": {
    "providers": {
      "clawbrain": {
        "baseUrl": "https://api.factorhub.cn/v1",
        "apiKey": "你的 API Key",
        "api": "openai-completions",
        "models": [
          { "id": "clawbrain-auto", "name": "ClawBrain Auto", "input": ["text", "image"], "contextWindow": 262144, "maxTokens": 65536 },
          { "id": "clawbrain-pro", "name": "ClawBrain Pro", "input": ["text", "image"], "contextWindow": 262144, "maxTokens": 65536 },
          { "id": "clawbrain-max", "name": "ClawBrain Max", "input": ["text", "image"], "contextWindow": 262144, "maxTokens": 65536 },
          { "id": "clawbrain-flash", "name": "ClawBrain Flash", "contextWindow": 262144, "maxTokens": 65536 }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": { "primary": "clawbrain/clawbrain-auto" }
    }
  }
}

然后重启:

openclaw gateway restart

获取 API Key

  1. 前往 clawbrain.dev/dashboard
  2. 注册账号(免费 50 次对话/天)
  3. 复制 API Key 到配置中

定价

方案 价格 每天对话次数
免费 ¥0 50
Pro ¥99/月 1,000
Pro Max ¥199/月 3,000
企业版 ¥299/月 无限

更多信息:clawbrain.dev

安全使用建议
Before installing: 1) Verify the provider domains and ownership — SKILL.md links to clawbrain.dev but the API baseUrl is api.factorhub.cn; confirm these are legitimately related. 2) Understand data flow: installing as described will send agent requests (and likely extracted 'memories') to the remote provider — read its privacy/data-retention policy and confirm what it stores and for how long. 3) Treat the API Key as sensitive: only use a provider you trust and avoid using it with sensitive or production data until you confirm behavior. 4) Because the skill is instruction-only, ask the maintainer (or vendor page) how features like DreamTask (daily consolidation), identity auto-sync, and 3D visualization are implemented and where they run (local vs. remote). 5) If unsure, do a limited test: don’t set it as your default model, and test with non-sensitive conversations to observe requests and responses before full adoption.
功能分析
Type: OpenClaw Skill Name: clawbrain-boost Version: 1.6.0 The skill bundle is a configuration and instruction guide for integrating the 'ClawBrain' LLM service into OpenClaw. It contains no executable code, instead providing manual setup steps for a third-party API provider (api.factorhub.cn) and behavioral guidelines for the agent, such as improving response accuracy and maintaining a memory system. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found.
能力标签
requires-sensitive-credentials
能力评估
Purpose & Capability
The description promises a rich local enhancement (memory system, scheduled consolidation tasks, 3D visualization, automatic identity sync). However the package is instruction-only and provides no code or install artifacts to implement local agents or schedulers. That could be fine if those features are provided by the remote model/service, but the skill metadata failed to declare the API credential it in practice requires. Also the SKILL.md points users to clawbrain.dev for an API key but the model baseUrl is 'https://api.factorhub.cn/v1' — a domain mismatch that is unexplained.
Instruction Scope
Instructions tell the user to edit ~/.openclaw/openclaw.json to add a remote provider and API key and to restart the gateway. Those actions are within scope for integrating a remote model provider. Concerns: SKILL.md asserts daily background tasks (DreamTask at 3:00), identity auto-sync, and local visualization without providing local components or explaining whether these run on the remote service. The instructions are otherwise specific and do not request unrelated system files or arbitrary secrets.
Install Mechanism
This is instruction-only with no install spec and no code files, so nothing is written or executed locally by the skill itself. That lowers local install risk. The README suggests using 'clawhub install clawbrain-boost' but there is no install artifact provided for review.
Credentials
Registry metadata lists no required environment variables or primary credential, but the runtime instructions explicitly require an API Key (to be stored in ~/.openclaw/openclaw.json) for a remote provider. That credential request is reasonable for a model-provider integration — but it should have been declared. The mismatch in declared vs. required credentials reduces transparency. Also the API endpoint domain (api.factorhub.cn) differs from the visible homepage (clawbrain.dev), which raises questions about where data (including conversation content/memory) will be sent and stored.
Persistence & Privilege
The skill is not always-included and model invocation is allowed (default). The SKILL.md tells users to change their OpenClaw default model to the remote provider, which is a normal configuration change but has the effect of routing agent requests (and potentially memory) to the external service. This is expected for provider integrations but is a behavioral/privilege change you should be aware of.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install clawbrain-boost
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /clawbrain-boost 触发
  4. 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
v1.6.0
v1.6: 品牌重新定位。智能路由→智能适配,四档模型→四档性能,Credits→次对话。新增数据保真能力描述。
v1.2.0
记忆可靠性升级+长对话支持+降级透明
v1.0.1
v1.0.1: 知识图谱替代文件系统 + DreamTask 自动整合 + 修正 API 地址和 contextWindow
v0.9.2
v0.9.2: 一键配置主动思考+记忆+输出验证+智能容错
元数据
Slug clawbrain-boost
版本 1.6.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 4
常见问题

Clawbrain Boost 是什么?

ClawBrain Boost v1.6 — 一键让 OpenClaw 更准更稳。记忆系统 + 数据保真 + 自动容错 + 写作助手。越用越懂你。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 138 次。

如何安装 Clawbrain Boost?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install clawbrain-boost」即可一键安装,无需额外配置。

Clawbrain Boost 是免费的吗?

是的,Clawbrain Boost 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Clawbrain Boost 支持哪些平台?

Clawbrain Boost 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 Clawbrain Boost?

由 michaelfeng(@michaelfeng)开发并维护,当前版本 v1.6.0。

💬 留言讨论