← Back to Skills Marketplace
michaelfeng

ClawBrain Doctor

by michaelfeng · GitHub ↗ · v1.2.0 · MIT-0
cross-platform ⚠ suspicious
174
Downloads
0
Stars
0
Active Installs
7
Versions
Install in OpenClaw
/install clawbrain-pro-doctor
Description
诊断你的 OpenClaw 配置和运行状态,基于 v1.2 的输出验证、模型健康监控、知识图谱检查、记忆来源标注、降级通知和长对话截断诊断
README (SKILL.md)

ClawBrain Doctor

检查你的 OpenClaw 配置和运行状态,告诉你哪里有问题、哪里可以优化。

使用方法

说"诊断一下"或"检查配置"。

常见问题和解决方案

问题:AI 让它做事,它跟你聊天

  • 原因:模型不习惯调用工具
  • 方案:ClawBrain Pro 的编排引擎会强制引导模型调用工具
  • v1.0 检测:链条完成强制(Chain Completion Enforcement),开始了就必须做完

问题:出了错就放弃

  • 原因:模型不知道该怎么恢复
  • 方案:ClawBrain Pro 自动注入恢复策略,甚至请另一个模型帮忙
  • v1.0 检测:三级恢复策略 + 双模型共识机制

问题:多步任务做一半就停了

  • 原因:模型不会规划
  • 方案:ClawBrain Pro 的编排引擎先规划步骤,再监督执行
  • v1.0 检测:主动思考框架(三问:不知道什么/可能出问题/怎么验证)

问题:听不懂模糊的话

  • 原因:模型不会联系上下文
  • 方案:ClawBrain Pro 自动分析意图,注入"先查什么再做什么"的提示
  • v1.0 检测:多维闭环(上游→下游→平行→全局一致性检查)

问题:回答质量不稳定

  • 原因:单模型没有自检能力
  • 方案:独立模型四维评分(准确性/完整性/逻辑性/格式),低于 70 分自动重来
  • v1.0 新增:输出验证器(Output Verifier)

问题:某个后端模型突然变慢或出错

  • 原因:模型服务不稳定
  • 方案:模型健康监控,5 次失败自动熔断 60s,半开探测后恢复
  • v1.0 新增:GET /v1/health/backends 查看实时状态

问题:记忆丢失或不准

  • 原因:知识图谱需要检查
  • 方案:查看图谱统计,清理过时实体
  • v1.0 新增:GET /v1/graph/stats 检查图谱健康度

问题:记忆来源标注不正确

  • 原因:记忆条目缺少来源引用或引用了错误的对话
  • 方案:检查记忆来源标注是否与原始对话匹配,清理错误标注
  • v1.2 新增:记忆来源标注诊断

问题:降级通知没有工作

  • 原因:记忆服务或后端模型降级时用户未收到通知
  • 方案:检查降级通知机制是否正常触发,确认用户能看到降级提示
  • v1.2 新增:降级通知诊断

问题:长对话截断后丢失上下文

  • 原因:超长对话被截断时关键信息未从记忆恢复
  • 方案:检查长对话截断时是否正确从知识图谱恢复关键上下文
  • v1.2 新增:长对话截断诊断

模型选择建议

你主要做什么 最适合的方案
简单文件/搜索 ClawBrain Flash(快且省)
复杂多步任务 ClawBrain Auto(自动升档)
经常出错要手动修 ClawBrain Pro(编排 + 错误恢复)
深度推理/代码生成 ClawBrain Max(深度思考)
模糊指令理解 ClawBrain Pro / Auto(上下文增强)

健康检查命令

# 检查后端模型健康状态
curl https://api.factorhub.cn/v1/health/backends -H "Authorization: Bearer YOUR_KEY"

# 检查知识图谱状态
curl https://api.factorhub.cn/v1/graph/stats -H "Authorization: Bearer YOUR_KEY"

# 检查模型列表和能力
curl https://api.factorhub.cn/v1/models -H "Authorization: Bearer YOUR_KEY"

接入 ClawBrain:https://clawbrain.dev/dashboard

Usage Guidance
This skill is inconsistent: it promises local diagnostics but only shows remote API checks and example curl commands that require an API key that the skill does not declare. Before installing or running it: 1) verify the operator/domain (api.factorhub.cn and clawbrain.dev) and look for an official homepage/source repository; 2) do not paste or expose your API key to an unknown skill — prefer running the shown curl commands yourself from a trusted shell; 3) ask the author to document required credentials (e.g., OPENCLAW_API_KEY) and to explain how the skill will access local agent configs/logs if it truly diagnoses runtime state; 4) if you need local diagnostics, prefer a skill that explicitly requests read access to specific config/log paths or provides reproducible local commands. Proceed only if you trust the external endpoints and the skill author and after they fix the credential/documentation gaps.
Capability Assessment
Purpose & Capability
The skill promises comprehensive 'diagnose OpenClaw configuration and runtime' but the SKILL.md only offers high-level guidance and three curl calls to api.factorhub.cn. It does not show any mechanism to read local OpenClaw configs, logs, or runtime state—so its actual capability is narrower than its description suggests.
Instruction Scope
Runtime instructions instruct running curl against api.factorhub.cn endpoints (including an Authorization: Bearer header example) but do not explain how the agent obtains the API key. The instructions do not reference any local files/configs, so they cannot perform the on-agent checks the description promises. The external endpoints are not documented (unknown domain).
Install Mechanism
Instruction-only skill with no install spec and only requires curl (declared). This is low-risk from install/execution perspective.
Credentials
SKILL.md examples use an Authorization: Bearer token, but the skill declares no required environment variables or primary credential. This mismatch—example commands needing an API key while the skill doesn't request or document how that key is provided—is a proportionality and transparency concern.
Persistence & Privilege
always:false and no install steps; the skill does not request persistent presence or elevated system privileges.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install clawbrain-pro-doctor
  3. After installation, invoke the skill by name or use /clawbrain-pro-doctor
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.2.0
新增3个诊断项:来源标注+降级通知+截断
v1.0.2
Fix display name
v1.0.1
v1.0.1: 新增输出验证器 + 模型健康熔断 + 知识图谱检查诊断
v0.9.4
统一品牌名
v0.9.3
更新 summary 描述
v0.9.2
v0.9.2: 更新诊断规则
v1.0.0
Initial release of clawbrain-pro-doctor. - Diagnose configuration issues with OpenClaw and provide optimization suggestions based on 205 scenario assessments. - Offers troubleshooting tips for common model behavior problems, such as failed multi-step tasks, poor intent recognition, and error recovery. - Guides users on using the diagnostic tool and recommends best-fit solutions for different usage scenarios. - Easily invoked by commands like "诊断一下" or "检查配置".
Metadata
Slug clawbrain-pro-doctor
Version 1.2.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 7
Frequently Asked Questions

What is ClawBrain Doctor?

诊断你的 OpenClaw 配置和运行状态,基于 v1.2 的输出验证、模型健康监控、知识图谱检查、记忆来源标注、降级通知和长对话截断诊断. It is an AI Agent Skill for Claude Code / OpenClaw, with 174 downloads so far.

How do I install ClawBrain Doctor?

Run "/install clawbrain-pro-doctor" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is ClawBrain Doctor free?

Yes, ClawBrain Doctor is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does ClawBrain Doctor support?

ClawBrain Doctor is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created ClawBrain Doctor?

It is built and maintained by michaelfeng (@michaelfeng); the current version is v1.2.0.

💬 Comments