Token导航 LogoToken导航TokenDH.com
研究检索操作浏览器github未标认证来源可访问许可证需确认审计通过

mole-debug摩尔调试

Agent Skill

mole-debug 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。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

776

周安装

32

GitHub Stars

4

下载量

253
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

2

许可证

unknown

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:mole-debug(摩尔调试)
来源仓库:https://github.com/autumnsgrove/groveengine
仓库路径:skills/mole-debug
安装命令:
npx skills add https://github.com/autumnsgrove/groveengine --skill mole-debug
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。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/autumnsgrove/groveengine --skill mole-debug

简介

用于查找、检索和筛选相关信息。mole-debug 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

  • 适合根据关键词或任务场景快速定位候选结果。
  • 可结合来源仓库和 README 核验具体用法。
  • 安装前建议确认权限范围和是否会触发联网或文件操作。
  • 注意维护状态,避免使用已废弃或不受支持的技能。

SKILL.md

The Mole ⛏️

The mole works in the dark. It doesn't need to see — it feels. Vibrations travel through the earth, and the mole follows them to their source with absolute patience. It never guesses. It never panics. It digs methodically, one tunnel at a time, testing the soil ahead before committing.

When to Activate

  • User says "debug this," "why is this failing," "something's broken," or "find this bug"
  • User calls /mole-debug or mentions mole/debug/investigate/diagnose
  • Tests are failing and the cause isn't obvious
  • A bug was reported but nobody can locate it
  • "It works locally but fails in production" or "it works on my machine"
  • Intermittent failures, flaky tests, race conditions
  • Error messages that don't point to the actual problem

IMPORTANT: The Mole ALWAYS reproduces first. If you can't make it fail on demand, you haven't started debugging — you're just guessing.

Pair with: beaver-build for regression tests after fix, panther-strike when the bug is already located, bloodhound-scout for understanding unfamiliar code paths, eagle-architect when the Three-Burrow Threshold triggers


The Dig

FEEL → DIG → TUNNEL → SURFACE → SEAL
  ↓       ↓       ↓        ↓        ↓
Repro-  Trace   Hypo-    Fix     Prevent
duce    Data    thesis   with    Recur-
Issue   Flow    Test     Test    rence

Phase 1: FEEL

*The mole presses its paw against the earth. Something trembles beneath...*

Reproduce the issue. Get it to fail on demand. No reproduction, no debugging.

Reproduction checklist:

  1. Exact error — Full error message, stack trace, HTTP status code. Not a summary.
  2. Exact steps — What sequence triggers the failure? Write them down.
  3. Exact environment — Node version, browser, OS, local vs production, which tenant.
  4. Minimal reproduction — Strip away everything unnecessary. Find the smallest case that still breaks.

If you CANNOT reproduce: Widen the search (different browsers, tenants, data states). Check for race conditions. Check git history: git bisect narrows the window.

Record the vibration in a Vibration Log (see references/debugging-strategies.md for template).

Output: A reliably reproducible failure, or a clear statement of what blocks reproduction.


Phase 2: DIG

*The mole begins to dig. Not frantically — deliberately...*

Trace the data flow from entry to exit. Instrument the boundaries.

Place sensors at strategic points — boundaries between systems, not scattered throughout code. Use [MOLE:LOCATION] labels so you can find and remove them later.

# Find the entry point
gf --agent func "functionName"
gf --agent search "POST /api/endpoint"

# Trace imports and impact
gf --agent usage "ServiceName"
gf --agent impact "src/lib/server/services/affected-file.ts"

For Grove-specific debugging patterns (routes, D1, KV, auth, build), see references/grove-debugging-patterns.md.

For detailed instrumentation techniques, see references/debugging-strategies.md.

Output: A map of the data flow with instrumentation at key boundaries.


Phase 3: TUNNEL

*The mole chooses a direction. One tunnel. One hypothesis...*

Form hypotheses. Test ONE at a time. Binary search the problem space.

CRITICAL: Never change two things between test runs. One variable at a time. Binary search is logarithmic — for a 16-layer call chain, it finds the fault in 4 steps.

Track every hypothesis in the Vibration Log:

Hypothesis 1: [description]
Test: [what you did]
Result: [CONFIRMED / REFUTED — why]

The Three-Burrow Threshold

HARD RULE: After 3 failed fix attempts in different locations, STOP.

Three failures in three different places = architectural problem, not isolated bug. Surface and call eagle-architect for an aerial view. Document everything in the Vibration Log.

Output: Either a confirmed root cause, or a Three-Burrow escalation with full documentation.


Phase 4: SURFACE

*The mole breaks through to daylight...*

Write a failing test that demonstrates the bug. Then fix the code. Then the test passes.

This order is mandatory:

1. Write test that reproduces the bug → TEST FAILS (proves bug exists)
2. Fix the code                       → TEST PASSES (proves fix works)
3. Run full affected CI               → ALL TESTS PASS (proves nothing else broke)
pnpm install
gw ci --affected --fail-fast --diagnose

Output: A passing test that proves the bug is fixed, and clean CI.


Phase 5: SEAL

*The mole packs the earth behind it, sealing the tunnel forever...*

Prevent recurrence.

  1. Regression test — Already done in SURFACE. Lives permanently in CI.
  2. Root cause comment — Brief comment at fix site explaining WHY, not WHAT.
  3. Class-of-bug analysis — Could this happen elsewhere? Search for similar patterns:
# If schema/query mismatch, check ALL queries
gf --agent search "SELECT.*content.*FROM posts"
# If missing tenant scope, check ALL tenant queries
gf --agent search "FROM posts WHERE"  # Missing tenant_id?
  1. Clean up — Remove ALL [MOLE:*] debug logging. The mole seals its tunnels.
gf --agent search "MOLE:"

Output: Sealed tunnel. Regression test. Root cause documented. Class-of-bug checked. Debug instrumentation removed.


Quick Decision Guide

SymptomApproach
Test fails, error is clearSkip to SURFACE — write the fix test, fix it
Test fails, error is crypticFEEL (reproduce) then DIG (trace data flow)
"Works on my machine"FEEL (compare environments), DIG (check configs)
Intermittent failureFEEL (find trigger pattern), DIG (state snapshots)
"It used to work"Time-travel: git bisect to find breaking commit
500 error, no useful messageDIG (instrument boundaries), check server logs
Multiple tests failingCheck common dependency — may be architectural
"Everything is broken"Three-Burrow likely — escalate to Eagle immediately

MUST DO

  • Reproduce the bug before attempting any fix
  • Write a failing test before writing the fix
  • Test one hypothesis at a time
  • Use the Vibration Log to track all attempts
  • Binary search the problem space
  • Remove all debug instrumentation before committing
  • Check for the same class of bug elsewhere
  • Run gw ci --affected --fail-fast --diagnose before considering complete
  • Respect the Three-Burrow Threshold

MUST NOT

  • Scatter console.logs randomly (instrument boundaries strategically)
  • Change multiple variables between test runs
  • Fix the symptom instead of the root cause
  • Skip reproduction ("I think I know what's wrong")
  • Leave debug logging in the codebase
  • Commit a fix without a regression test
  • Dig past the Three-Burrow Threshold without escalating

References

  • references/grove-debugging-patterns.md — SvelteKit/Cloudflare-specific patterns
  • references/debugging-strategies.md — Detailed techniques (binary search, boundary instrumentation, state snapshots, git bisect, hypothesis tracking, Vibration Log template)

*The earth is quiet now. The vibration has stopped. Somewhere beneath the grove, a sealed tunnel marks where the mole found the truth.* ⛏️

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

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

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

Codex

36.05%
按下载量换算91

Claude

27.89%
按下载量换算71

Cursor

19.32%
按下载量换算49

Gemini CLI

8.29%
按下载量换算21

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

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。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills