项目真源侦探|接手不走错目录

SKILL开发
开发 未标注权限 许可证不明确
安装这个技能

拆开物理目录、项目身份、Git 真源和真实运行态,识别旧副本、worktree 与构建产物,并给出可信接手入口。

一句话选型结论

按资料字段整理

项目真源侦探|接手不走错目录是开发类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类开发 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_59dd9ccb/project-takeover-truth-audit
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

开发
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_59dd9ccb仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验未标注上游最后核验时间 · 数据更新 1791621870
正文体量5.5 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
写代码与改 bug

按需求补代码、定位并修掉报错。

读仓库与生成改动

让模型读懂项目结构,再给出可落地的改动。

跑测试与排查故障

按现象排查问题,补测试用例并复跑。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在开发场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
项目真源侦探|接手不走错目录当前 未标注权限 未标注 未标注 未标注 有
技能审查员 浏览器控制 ClawHub 市场 877.9 万 OpenClaw 有
Vercel React 最佳实践 浏览器控制 GitHub 仓库 849.1 万 Codex、Claude、Cursor 有
GitHub 代码协作 联网访问 ClawHub 市场 763.3 万 OpenClaw 有
网页设计指南 联网访问 GitHub 仓库 668.6 万 Codex、Claude、Cursor 有
本体论 只读 ClawHub 市场 663.7 万 OpenClaw 有

SKILL.md 全文

5.5 KB
模型实际读到的说明书原文,来自技能仓库

项目真源侦探|接手不走错目录

Determine what the project is, which copy is authoritative, and what is actually running before recommending any continuation point.

Keep four layers separate

Never collapse these into one claim:

  1. Physical directory: a folder that exists.
  2. Project identity: the product or codebase that folder represents.
  3. Repository lineage: Git root, remote, branch, HEAD, worktree, clone, archive, or extracted ZIP relationship.
  4. Live surface: the process, listener, build output, deployed URL, or app currently serving users.

downloaded, copied, enabled, running, deployed, and continued from are different states.

Safety boundary

  • Start read-only. Do not modify files or Git state.
  • Do not read or print .env, tokens, cookies, keychains, private databases, session stores, or user data.
  • Do not run git pull, checkout, reset, clean, install commands, process restarts, or deployment actions.
  • Restrict discovery to the explicit target and the smallest relevant parent or sibling set. Do not recursively scan a home directory or disk by default.
  • Preserve dirty worktrees and unknown files. Report them; do not clean them.

Audit sequence

  1. Resolve the exact target path and scope. If the user names several similar copies, keep them separate until evidence proves a relationship.
  2. Read authority files in order when present: current user instruction, SOURCE_OF_TRUTH.md, recent status or handoff, AGENTS.md, README, package metadata, then code and runtime evidence. If root markers do not settle the question, search only likely documentation locations inside the target for focused names such as handoff, status, target-install, source, or deploy; do not crawl unrelated trees.
  3. Run the bundled read-only snapshot:
python3 scripts/inspect_project.py /absolute/path/to/project
  1. Verify Git lineage with root, remotes, branch, HEAD, upstream, shallow state, worktrees, submodules, and dirty status. A matching name is not proof of shared lineage.
  2. Identify source/build/deploy boundaries. Treat dist, build, out, exported static pages, archives, and generated folders as outputs unless current evidence proves otherwise.
  3. If runtime matters, inspect the listener and its process command line. Match the actual executable arguments and working/source path to a candidate directory; do not infer from the browser URL alone.
  4. Compare suspected duplicates using remote URL, commit ancestry, project markers, recent timestamps, and small targeted hashes. Do not compare entire disks.
  5. Classify each claim as verified, inferred, conflicting, or unknown.
  6. Classify each candidate directory as authoritative source, active feature worktree, clone, copied snapshot, generated output, archive, or unknown. Use copied snapshot only when targeted content evidence or explicit source-to-target documentation establishes the direction.

Read references/evidence-matrix.md when multiple copies, worktrees, builds, or live processes are involved. Read references/report-template.md before answering.

Decision rules

  • Prefer explicit current authority documents and current runtime evidence over README prose or an old cached page.
  • A directory inside a Git repo is not automatically the repo root.
  • A worktree is part of one repository lineage, not a new project identity by itself.
  • An extracted ZIP or old clone is a separate physical copy until history and remotes prove lineage.
  • A clean worktree does not prove it is current; compare HEAD and remote only when network access is authorized and necessary.
  • A local remote-tracking ref such as origin/main proves only the last fetched state. Say live remote verified only after an authorized network refresh.
  • When the main worktree and an unmerged feature worktree both exist, give two decisions when relevant: the general project entry point and the feature-specific continuation point. Do not collapse them into one winner.
  • A successful build does not prove the built artifact is the live surface.
  • If evidence conflicts, report the conflict and the smallest next read-only check. Do not select a winner by intuition.

Answer contract

Lead with one takeover decision:

Continue from: <exact path or unable to confirm>
Why: <strongest evidence>
Do not continue from: <stale/duplicate/output path and reason>

Then show the identity map, evidence, risks, and next safe action. Use 无法确认 when proof is missing. Do not turn an audit request into implementation.

常见问题

5 问
答案由本站依据条目字段整理
项目真源侦探|接手不走错目录怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_59dd9ccb/project-takeover-truth-audit。安装方式为未标注。装完刷新技能列表或重开客户端。

项目真源侦探|接手不走错目录支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

项目真源侦探|接手不走错目录权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

项目真源侦探|接手不走错目录收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

项目真源侦探|接手不走错目录用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 未标注。