Token导航 LogoToken导航TokenDH.com
开发需要联网github未标认证来源可访问许可证需确认审计通过

investigate调查

Agent Skill

investigate 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。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

196

周安装

8

GitHub Stars

公开资料未说明

下载量

63
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:investigate(调查)
来源仓库:https://github.com/timlai666/skills
仓库路径:skills/investigate
安装命令:
npx skills add https://github.com/timlai666/skills --skill investigate
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/timlai666/skills --skill investigate

简介

用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息。

  • 适合围绕仓库状态、代码变更或协作事项进行整理。
  • 可结合来源仓库和原始 README 继续核验具体用法。
  • 安装前建议确认权限范围、维护状态及是否会触发联网或命令执行。
  • 涉及文件读写时应先明确输入输出范围。investigate 属于开发类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Preamble (run first)

_BRANCH=$(git branch --show-current 2>/dev/null || echo "unknown")
echo "BRANCH: $_BRANCH"
# Auto-activate freeze to prevent accidental changes during investigation
_FREEZE=$(cat ~/.mystack/freeze.txt 2>/dev/null || echo "none")
if [ "$_FREEZE" = "none" ]; then
  echo "." > ~/.mystack/freeze.txt
  echo "FREEZE_AUTO_ACTIVATED: edits locked to current directory during investigation"
fi

Note: Freeze is auto-activated during investigation to prevent accidentally "fixing" unrelated code. Run /stacksmith-safety off after investigation to remove.


Iron Law

No fix without understanding the root cause.

If you have attempted 3 different fixes and none worked: STOP and escalate. Do not guess a 4th fix. More attempts without new information just waste time and may make the system harder to understand.


Step 1??Understand the symptom

Ask if not clearly provided:

  • Exact symptom??error message verbatim, wrong output, crash, silent failure?
  • Reproduction??always, sometimes (what triggers it), or intermittent?
  • When did it start??after what change? Or has it always been this way?
  • Environment??local, staging, prod? All three?
  • What was tried??what has already been attempted?

Step 2??Form hypotheses

Based on the symptom, list 3?? possible root causes ranked by likelihood. For each, state the evidence that points to it:

H1 (most likely, ~60%): [hypothesis]
  Evidence: [why you think this]
  Rules out: [what would disprove it]

H2 (~25%): [hypothesis]
  Evidence: [why you think this]
  Rules out: [what would disprove it]

H3 (~10%): [hypothesis]
  Evidence: [why you think this]

H4 (~5%): [hypothesis]

Common hypothesis categories:

  • Wrong assumption about data (nil where non-nil expected, wrong type, wrong shape)
  • State/timing issue (race condition, stale cache, wrong ordering)
  • Config/environment mismatch (works local, fails staging)
  • External dependency changed (API, library version)
  • Regression from recent change (git log --oneline -10)

Step 3??Test hypotheses cheaply (observe before touching code)

For each hypothesis, design a minimal test that confirms or rules it out??without modifying code:

HypothesisTest methodCost
H1: value is nilAdd puts/console.log/print at boundaryTrivial
H2: race conditionLog timestamps of competing operationsTrivial
H3: config issuePrint ENV values at startupTrivial
H4: library bugPin version and reproduceLow

Run the cheapest tests first. Report results. Eliminate hypotheses.

# Check recent changes that might be relevant
git log --oneline -15
git log --since="7 days ago" --name-only --format="" | sort | uniq -c | sort -rn | head -10

Step 4??Trace the data flow

Once narrowed to 1?? hypotheses, trace the exact path of the bad data:

[Request/Event enters at: ]
  ??[Layer 1] ??state here: [what you found]
  ??[Layer 2] ??state here: [what you found]
  ??[Where behavior diverges from expected] ??state here: [what you found]
  ??[Where the symptom manifests]

Identify the exact file:line where the wrong thing happens. This is the root cause.


Step 5??Fix (only after root cause confirmed)

Minimal fix??change only what's necessary to address the root cause, not the symptom.

Before writing the fix, state:

  • "Root cause: [exact description]"
  • "Fix: [what will change and why this addresses the root cause, not just the symptom]"
  • "Does not fix: [what this won't change]"
git add -p   # stage only the fix
git commit -m "fix: [root cause in plain English]\n\nInvestigation: [what was found and where]"

Write a regression test immediately after:

  • The test must fail before the fix and pass after
  • Tests the root cause, not the symptom
  • Commit: git commit -m "test: regression for [root cause]"

Step 6??Verify

Reproduce the original symptom after the fix. Confirm it's gone.

If the symptom persists: you may have fixed a symptom, not the root cause. Go back to Step 2 with updated hypotheses.


Step 7??Escalation (after 3 failed attempts)

If three different fixes have been tried and none resolved the symptom:

??ESCALATION ??3 attempts failed

Attempts:
1. [Fix 1] ??what changed ??result: [what happened]
2. [Fix 2] ??what changed ??result: [what happened]
3. [Fix 3] ??what changed ??result: [what happened]

Remaining hypotheses:
H?: [hypothesis] ??need to test: [what information/access/expertise is required]

Current best understanding:
The root cause is likely in [area], but I cannot verify without [specific missing info].

Recommended next step:
[Specific action: add more logging, check external service, get access to prod logs, pair with someone who knows the system]

STOP HERE. Do not attempt a 4th fix.


Log + unfreeze reminder

echo "{\"ts\":\"$(date -u +%Y-%m-%dT%H:%M:%SZ)\",\"skill\":\"investigate\",\"branch\":\"$(git branch --show-current 2>/dev/null || echo 'N/A')\",\"outcome\":\"success\",\"repo\":\"$(basename $(git rev-parse --show-toplevel 2>/dev/null) 2>/dev/null || echo 'N/A')\"}" >> ~/.mystack/timeline.jsonl

After investigation completes: "Investigation complete. Auto-freeze is still active. Run /stacksmith-safety off to remove edit restrictions."

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

32.38%
按下载量换算20

Claude

31.92%
按下载量换算20

Cursor

18.79%
按下载量换算12

Gemini CLI

9.33%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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