ClawBrain Pro

ClawBrain v1.6 — 更准、更稳、越用越懂你的大模型。记忆系统 + 数据保真 + 自动容错 + 输出验证。21 场景评测第一。

Security Scan

The skill appears internally consistent with a model-integration/agent-augmentation purpose (memory, retry, verification) and only instructs the agent to configure and call an external ClawBrain API, but pay attention to where your API key and memory data will be sent before installing.

Purpose & Capability

The name/description (agent memory, error-retry, output verification, orchestration) match the instructions: SKILL.md shows how to add ClawBrain as a model provider, edit ~/.openclaw/openclaw.json, run clawhub/openclaw commands, and call api.factorhub.cn and clawbrain.dev. Minor inconsistency: the skill metadata declares no required env vars, yet the docs repeatedly instruct the user to provide an API Key in config and examples show Authorization headers — functionally normal but not declared as a required environment credential in the metadata. Also the documentation references multiple domains (api.factorhub.cn, clawbrain.dev, and GitHub repo) which is plausible but worth verifying.

Instruction Scope

The runtime instructions stay within the stated purpose: they show configuration snippets, curl examples for health/models endpoints, and commands to install/enable the integration (clawhub install, openclaw gateway restart). Sub-skill files include curl-based health checks and mention scheduled daily memory consolidation. The instructions do require network access and an external API key (expected for a third-party model service), but they do not direct the agent to read unrelated system secrets or transmit data to unexpected endpoints beyond the documented service domains.

Install Mechanism

This is instruction-only (no install spec, no code files). That is low-risk: nothing is automatically downloaded or written by the skill itself. The SKILL.md expects the user to run external installers (clawhub) and edit config, which is user-driven and not performed by the skill package.

Credentials

No required environment variables or credentials are declared in the metadata, but the docs explicitly require an API Key in the OpenClaw config and show curl Authorization headers. Requiring a single provider API key is proportionate to the stated purpose, but the mismatch between declared requires and the instructions is worth noting. The skill does not request unrelated credentials or system config paths.

Persistence & Privilege

The skill does not set always:true and is user-invocable. It instructs the user to modify their OpenClaw configuration and restart the gateway (normal for adding a new provider). There is no indication the skill attempts to modify other skills' configs or demand elevated system privileges.

Assessment

This skill is an integration guide for a third‑party model service (ClawBrain). Before installing: 1) Verify the external domains (api.factorhub.cn, clawbrain.dev, and the referenced GitHub repo) are trustworthy and you are comfortable sending prompts/contexts to them. 2) Understand that you will place an API Key into your OpenClaw config (~/.openclaw/openclaw.json) — avoid reusing high‑privilege or unrelated credentials. 3) The skill’s sub-pages use curl and recommend scheduled memory consolidation (daily tasks) — confirm where long‑term memories are stored and how they are protected/retained by the provider. 4) If you want to limit risk, create a restricted test API key/account and inspect network traffic or logs during initial use. 5) The metadata omits declaring the API key requirement even though the docs rely on it; consider this a documentation mismatch and verify credentials and endpoints before proceeding.

给你的 OpenClaw 装上更准、更稳、越用越懂你的大脑。

它能帮你做什么

• 复杂任务自动规划：你说一句话，它拆成步骤按顺序执行
• 智能适配：自动判断任务类型，调整推理深度，给出最优回答
• 出错自动修复：文件找不到自己去翻，命令出错换方法试，失败自动切换策略
• 做事做到底：多步任务每一步都盯着，不半途而废
• 听得懂模糊的话：说"帮我准备下"，它知道先查什么再做什么
• 结果自动验证：独立四维评分（准确/完整/逻辑/格式），不合格就重来
• 记忆系统：跨会话记住你的偏好、项目、决策，越用越懂你
• 数据保真：你的数字/日期/人名不会被篡改，生成后逐一校验
• 记忆有来源：清晰区分"你说过的原话"和"AI 归纳的摘要"
• 身份实时更新：告诉 AI "我换工作了"，它马上更新
• 长对话不崩溃：超长对话自动智能压缩而非报错
• 兼容 Claude Code：支持 Anthropic Messages API（/v1/messages）

评测数据（21 场景实测）

能力	ClawBrain Auto	ClawBrain Pro	最好的通用模型
综合得分	90%	86%	83%
错误恢复	100%	100%	80%
模糊指令	100%	100%	65%
多步任务	80%	80%	80%

接入方法

{
  "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 openclaw gateway restart。

其他工具

clawhub install clawbrain-boost            # 一键优化配置 + 记忆 + SOUL

获取 API Key：https://clawbrain.dev/dashboard