Token导航 LogoToken导航TokenDH.com
研究检索external-servicegithub未标认证来源可访问clear审计异常

error-debugger错误调试器

Agent Skill

error-debugger 用于记录任务执行中的错误、用户纠正、经验和能力缺口,适合在 Codex、Claude、Cursor、Gemini CLI 中希望让 Agent 持续沉淀问题、修正和最佳实践时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

973

周安装

39

GitHub Stars

14

下载量

315
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

3

许可证

MIT

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:error-debugger(错误调试器)
来源仓库:https://github.com/jackspace/claudeskillz
仓库路径:skills/error-debugger
安装命令:
npx skills add https://github.com/jackspace/claudeskillz --skill error-debugger
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/jackspace/claudeskillz --skill error-debugger

简介

error-debugger 用于记录任务执行中的错误与修正经验,帮助 Agent 持续优化问题处理流程。

  • 适用于 Codex、Claude、Cursor、Gemini CLI 中需要自动沉淀错误模式、提升调试效率的场景。
  • 通过分析错误信息、过往解决方案和回归测试,提供即时修复建议并保存至记忆库。
  • 安装方式:通过 npx 从 GitHub 仓库添加,需确认是否具备文件读写或命令执行权限。
  • 使用前请检查技能维护状态及是否涉及敏感操作,避免在生产环境直接运行未验证代码。

SKILL.md

Error Debugger

Purpose

Context-aware debugging that learns from past solutions. When an error occurs:

  1. Searches memory for similar past errors
  2. Analyzes error message and stack trace
  3. Provides immediate fix with code examples
  4. Creates regression test via testing-builder
  5. Saves solution to memory for future

For ADHD users: Eliminates debugging frustration - instant, actionable fixes. For SDAM users: Recalls past solutions you've already found. For all users: Gets smarter over time as it learns from your codebase.

Activation Triggers

  • User says: "debug this", "fix this error", "why is this failing"
  • Error messages containing: TypeError, ReferenceError, SyntaxError, ECONNREFUSED, CORS, 404, 500, etc.
  • Stack traces pasted into conversation
  • "Something's broken" or similar expressions

Core Workflow

1. Parse Error

Extract key information:

{
  error_type: "TypeError|ReferenceError|ECONNREFUSED|...",
  message: "Cannot read property 'map' of undefined",
  stack_trace: [...],
  file: "src/components/UserList.jsx",
  line: 42,
  context: "Rendering user list"
}

2. Search Past Solutions

Query context-manager:

search memories for:
- error_type match
- similar message (fuzzy match)
- same file/component if available
- related tags (if previously tagged)

If match found:

🔍 Found similar past error!

📝 3 months ago: TypeError in UserList component
✅ Solution: Added null check before map
⏱️ Fixed in: 5 minutes
🔗 Memory: procedures/{uuid}.md

Applying the same solution...

If no match:

🆕 New error - analyzing...
(Will save solution after fix)

3. Analyze Error

See reference.md for comprehensive error pattern library.

Quick common patterns:

  • TypeError: Cannot read property 'X' of undefined → Optional chaining + defaults
  • ECONNREFUSED → Check service running, verify ports
  • CORS errors → Configure CORS headers
  • 404 Not Found → Verify route definition
  • 500 Internal Server Error → Check server logs

4. Provide Fix

Format:

🔧 Error Analysis

**Type**: {error_type}
**Location**: {file}:{line}
**Cause**: {root_cause_explanation}

**Fix**:

// ❌ Current code const users = data.users; return users.map(user => <div>{user.name}</div>);

// ✅ Fixed code const users = data?.users || []; return users.map(user => <div>{user.name}</div>);


**Explanation**: Added optional chaining and default empty array to handle case where data or data.users is undefined.

**Prevention**: Always validate API response structure before using.

**Next steps**:

1. Apply the fix
2. Test manually
3. I'll create a regression test

5. Save Solution

After fix confirmed working:

# Save to context-manager as PROCEDURE
remember: Fix for TypeError in map operations
Type: PROCEDURE
Tags: error, typescript, array-operations
Content: When getting "Cannot read property 'map' of undefined",
         add optional chaining and default empty array:
         data?.users || []

Memory structure:

# PROCEDURE: Fix TypeError in map operations

**Error Type**: TypeError
**Message Pattern**: Cannot read property 'map' of undefined
**Context**: Array operations on potentially undefined data

## Solution

Use optional chaining and default values:

// Before const items = data.items; return items.map(...)

// After const items = data?.items || []; return items.map(...)


## When to Apply

- API responses that might be undefined
- Props that might not be passed
- Array operations on uncertain data

## Tested

✅ Fixed in UserList component (2025-10-17) ✅ Regression test: tests/components/UserList.test.jsx

## Tags

error, typescript, array-operations, undefined-handling

6. Create Regression Test

Automatically invoke testing-builder:


create regression test for this fix:

- Test that component handles undefined data
- Test that component handles empty array
- Test that component works with valid data

Tool Persistence Pattern (Meta-Learning)

Critical principle from self-analysis: Never give up on first obstacle. Try 3 approaches before abandoning a solution path.

Debugging Tools Hierarchy

When debugging an error, try these tools in sequence:

1. Search Past Solutions (context-manager)

# First approach: Check memory
search memories for error pattern

If no past solution found → Continue to next approach

2. GitHub Copilot CLI Search

# Second approach: Search public issues
copilot "Search GitHub for solutions to: $ERROR_MESSAGE"

If Copilot doesn't find good results → Continue to next approach

3. Web Search with Current Context

# Third approach: Real-time web search
[Use web search for latest Stack Overflow solutions]

If web search fails → Then ask user for more context

Real Example from Meta-Analysis

What happened: Tried GitHub MCP → Got auth error → Immediately gave up

What should have happened:

  1. Try GitHub MCP → Auth error
  2. Try gh CLI → Check if authenticated
  3. Try direct GitHub API → Use personal token
  4. Then create manual instructions if all fail

Outcome: The gh CLI WAS authenticated and worked perfectly. We gave up too early.

Applying This to Error Debugging

When fixing an error:

// Pattern: Try 3 fix approaches
async function debugError(error) {
  // Approach 1: Past solution
  const pastFix = await searchMemories(error);
  if (pastFix?.success_rate > 80%) {
    return applyPastFix(pastFix);
  }

  // Approach 2: Pattern matching
  const commonFix = matchErrorPattern(error);
  if (commonFix) {
    return applyCommonFix(commonFix);
  }

  // Approach 3: External search (Copilot/Web)
  const externalSolution = await searchExternalSolutions(error);
  if (externalSolution) {
    return applyExternalSolution(externalSolution);
  }

  // Only NOW ask for more context
  return askUserForMoreContext(error);
}

Integration Tool Persistence

When integrations are available, use them in this order:

For Error Search:

  1. GitHub Copilot CLI → Search issues in your repos and similar projects
  2. Local memory → Past solutions you've saved
  3. Web search → Latest Stack Overflow/docs

For Solutions:

  1. Past solution from memory (fastest)
  2. Codegen-ai agent (if complex bug) → Automated PR
  3. Jules CLI async task (if time-consuming fix)
  4. Manual fix with code examples

Metrics

Track debugging approach success:

{
  "error_id": "uuid",
  "approaches_tried": [
    {"type": "memory_search", "result": "no_match"},
    {"type": "copilot_search", "result": "success", "time": "5s"},
    {"type": "applied_fix", "verified": true}
  ],
  "total_time": "30s",
  "lesson": "Copilot found solution on second try"
}

Key insight: Most "failed" approaches are actually "didn't try enough" approaches.

Context Integration

Query Past Solutions

Before analyzing new error:

// Search context-manager
const pastSolutions = searchMemories({
  type: 'PROCEDURE',
  tags: [errorType, language, framework],
  content: errorMessage,
  fuzzyMatch: true
});

if (pastSolutions.length > 0) {
  // Show user the past solution
  // Ask if they want to apply it
  // If yes, apply and test
  // If no, analyze fresh
}

Learning Over Time

Track which solutions work:

{
  solution_id: "uuid",
  error_pattern: "TypeError.*map.*undefined",
  times_applied: 5,
  success_rate: 100%,
  last_used: "2025-10-15",
  avg_fix_time: "2 minutes"
}

Sort solutions by success rate when multiple matches found.

Project-Specific Patterns

Some errors are project-specific:

// BOOSTBOX-specific
Error: "Boost ID not found"
→ Solution: Check boost exists before processing

// Tool Hub-specific
Error: "Tool not installed"
→ Solution: Run tool installer first

// Save these as PROJECT-specific procedures

Integration with Other Skills

Testing Builder

After providing fix:

Automatically invoke: testing-builder
Create regression test for: {error_scenario}
Ensure test fails without fix, passes with fix

Context Manager

Query for similar errors:

search memories for:
- PROCEDURE type
- Error tag
- Similar message
- Same file/component

Save new solutions:

Save as PROCEDURE:
- Error pattern
- Solution
- Code examples
- Tested timestamp

Rapid Prototyper

For complex fixes:

If fix requires significant refactoring:
→ Invoke rapid-prototyper
→ Create isolated example showing fix
→ User validates before applying to codebase

Additional Resources

Quick Reference

Common Error Patterns

ErrorQuick Fix
undefined.map`data?.array
X is not a functionCheck function exists
ECONNREFUSEDCheck service running
CORSConfigure CORS headers
404Verify route exists
500Check server logs
TimeoutIncrease timeout value
Cannot find moduleInstall dependency

Trigger Phrases

  • "debug this"
  • "fix this error"
  • "why is this failing"
  • "something's broken"
  • [paste error message]
  • [paste stack trace]

File Locations

  • Past solutions: ~/.claude-memories/procedures/ (Linux/macOS) or %USERPROFILE%\.claude-memories\procedures\ (Windows)
  • Error patterns: Tagged with "error" in memory index

Success Criteria

✅ Common errors fixed instantly (<30 seconds) ✅ Past solutions automatically recalled ✅ All fixes include code examples ✅ Regression tests created automatically ✅ Solutions saved for future reference ✅ Debugging gets faster over time

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

04

需要参考平台分布和安装热度时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Claude Code

31.93%
按下载量换算101

windsurf

21.19%
按下载量换算67

OpenCode

17.07%
按下载量换算54

Codex

12.39%
按下载量换算39

Antigravity

6.93%
按下载量换算22

Gemini CLI

3.15%
按下载量换算10

安全审计

Gen Agent Trust Hub

未通过

Socket

通过

Snyk

可疑

权限和风险

external-service

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。

来源信息

继续浏览同类 Skills