Token导航 LogoToken导航

Agent Skill · 研究检索

archive-searcher(档案搜索器)

archive-searcher 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。

archive-searcher 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 60,488。

总安装 367周安装 15GitHub Stars 9下载 118核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:archive-searcher(档案搜索器)
来源仓库:https://github.com/yuniorglez/gemini-elite-core
仓库路径:skills/archive-searcher
安装命令:
npx skills add https://github.com/yuniorglez/gemini-elite-core --skill archive-searcher
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/yuniorglez/gemini-elite-core --skill archive-searcher

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于 Git 历史深度搜索和归档内容检索。支持 pickaxe 策略和 reflog 回溯等专业搜索技术。适用于历史提交分析和特征追溯场景。archive-searcher 属于研究检索类 Skill,可作为该场景下的辅助能力补充。安装需指定 GitHub 仓库,使用时需确认 Git 版本兼容性。

涉及历史数据访问时应注意隐私保护和合规要求。

来源仓库内路径:skills/archive-searcher

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/yuniorglez/gemini-elite-core --skill archive-s…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

archive-searcher 怎么安装?

在终端执行 npx skills add https://github.com/yuniorglez/gemini-elite-core --skill archive-searcher。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

archive-searcher 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

archive-searcher 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

archive-searcher 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

archive-searcher 用的人多吗?

公开数据是累计安装 367 次,近一周安装 15 次,下载 118 次,GitHub 9 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

archive-searcher 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
archive-searcher

档案搜索器 · 本页

3679接触敏感数据未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

📂 Skill: archive-searcher v2.0.0

Optimized for Git 2.48+ and Conductor Track Management (2026 Standards)

📋 Table of Contents

  1. Executive Summary
  2. Core Philosophy
  3. The Art of Archival Investigation
  4. Git Forensics: The Pickaxe Strategy (-S)
  5. Regex-Based Content Tracking (-G)
  6. Reflog: The Ultimate Safety Net
  7. Archival Search: Conductor Tracks
  8. Case Study: The Ghost of a Feature
  9. Git Internal Object Analysis
  10. Forensic Interviewing of Code
  11. Advanced Forensic Interviewing Scenarios
  12. Historical Contextualization
  13. Integrating with AI Analysis
  14. Comparison of Search Tools
  15. Best Practices for Searchability
  16. The "Do Not" List (Common Pitfalls)
  17. Step-by-Step Implementation Patterns
  18. Common Forensic Scenarios (2026 Edition)
  19. Mental Models for Forensic Excellence
  20. Advanced Troubleshooting & Recovery
  21. The Bisect Walkthrough: Automated Regressions
  22. Restoring Deleted Branches & History
  23. Searching for File Renames & Moves
  24. Archival Integrity & Verification
  25. Legal & Compliance Search (Audit Trails)
  26. Archival Security & Privacy
  27. Search Scenarios for 2026 AI Teams
  28. Integrating with External Search APIs
  29. Recovery from Corrupt Repositories
  30. Future-Proofing Your Archives
  31. Common Questions and Answers (FAQ)
  32. Glossary of Forensic Terms
  33. Tooling & Scripts
  34. Reference Documentation

🚀 Executive Summary

The archive-searcher skill is the definitive guide for high-precision historical analysis in modern software environments. It transforms the standard, often futile "grep" approach into a multi-layered, systematic forensic investigation. Whether you are hunting for a lost configuration from six months ago, tracing the exact commit that introduced a subtle race condition in a high-concurrency environment, or fulfilling a legal audit requirement, this skill provides the tactical commands and mental models needed to navigate deep history with absolute confidence.

In 2026, the landscape of software development has shifted. AI-driven code generation, automated refactors, and rapid experimentation have increased the "noise" in our repositories. The ability to reconstruct *why* a specific decision was made, what the constraints were at the time, and who the stakeholders were is now more valuable than the actual lines of code themselves. This skill focuses on both the "Git-Way" (raw historical data) and the "Conductor-Way" (structured mission context) of archiving and retrieving knowledge.

🧠 Core Philosophy

Information in a well-managed project is never truly "deleted"; it is simply "moved to the background." This skill is built upon four foundational pillars:

  • Persistence of Truth: Git history is an immutable ledger (in most cases) that contains the "Why" behind the "What." Every line of code, every deleted file, and every merge conflict is a snapshot of a human or AI decision. We treat the repository as a living organism where every scar tells a story.
  • Context Over Content: Finding the code is only Step 1. True archival search involves finding the surrounding commit message, the author, the timestamp, and the PR context. Without context, code is just text. We prioritize the "Meta-Data" of the change.
  • Surgical Efficiency: We prioritize indexed and intelligent searches (like git log -S) over brute-forcing history. This saves time and reduces cognitive load during high-pressure debugging sessions. Efficiency is not just about speed; it's about accuracy.
  • Layered Defense Strategy: If a piece of information isn't in the current branch, we check other branches. If it's not in the visible git log, we check git reflog. If it's completely missing from Git, we pivot to the conductor archives. We never stop at the first "no."

🎭 The Art of Archival Investigation

Investigating history is more like archaeology than engineering. You are looking for fragments of logic buried under layers of refactors.

  • Stratigraphy: Understanding which "layer" of the project you are looking at (e.g., the Pre-Next.js era vs. the Modern era).
  • Artifact Analysis: Treating every commit as an artifact that must be dated and attributed.
  • Reconstruction: Using partial data to rebuild the full intent of a feature.

🛠 Git Forensics: The Pickaxe Strategy

1. The -S (Pickaxe) Search

The Pickaxe is the most efficient way to find when a specific string was introduced or removed. Unlike a standard grep, it only shows commits where the *number of occurrences* of the string changed.

Why use it?

  • Lifecycle Tracking: To find when a function was first added and when it was finally retired.
  • Constant Hunting: To find when a specific constant, magic number, or hardcoded API key was deleted.
  • Variable Forensics: To trace how a specific variable name has migrated through the codebase over years of refactors.
# Find when 'DeprecatedAPI' was added or removed
# This will show you the exact commits where the usage count changed.
# The --patch flag is critical here to see the actual diff.
git log -S "DeprecatedAPI" --patch --oneline

Advanced Usage: Always use it with --all to search through every branch, including merged, orphaned, or abandoned ones.

git log -S "auth_token_v1" --all --graph --decorate --oneline

2. Regex-Based Content Tracking (-G)

While -S looks for changes in count, -G looks for changes that *match a regular expression*. This is much more powerful for finding structural changes or patterns that have many variations.

# Find commits where lines matching a regex were added/removed
# Example: Finding any change to a port number between 3000-3999
# This is perfect for security audits or infrastructure changes.
# Note: This is computationally more expensive than -S.
git log -G "port: 3[0-9]{3}" --patch

🕰 Reflog: The Ultimate Safety Net

git reflog is the "history of your history." It tracks every time the HEAD of a branch moves.

What Reflog captures that Log does not:

  • Commit Amends: If you git commit --amend, the original commit is still accessible.
  • Rebase Casualties: Pre-rebase commits are saved here.
  • Hard Resets: Rescues you from an accidental reset --hard.
  • Deleted Branches: Recovers the last known state of a deleted branch.
# View the reflog with relative dates for easier navigation
git reflog --date=relative

📂 Archival Search: Conductor Tracks

In the Squaads ecosystem, we use "Conductor" to manage mission tracks. This provides a semantic layer on top of raw Git commits.

Archival Hierarchy

  • conductor/archive/: Completed project-level archives.
  • conductor/tracks/<track_id>/archive/: Internal history of a specific track.
  • conductor/index.md: The central directory of all missions.

🕵️‍♂️ Case Study: The Ghost of a Feature

Problem: A "Team Collaboration Dashboard" was fully functional in Feb 2025. By Jan 2026, it is gone.

Detailed Investigation Walkthrough:

  1. Initial Search: git log --all --grep="Collaboration Dashboard"

- *Discovery*: Commit e5f1b2c (Oct 2025) says "Cleanup legacy UI".

  1. Blast Radius Analysis: git show e5f1b2c --stat

- *Discovery*: Massive deletion of components.

  1. Finding the Golden Version: git log --diff-filter=A -- src/components/Dashboard.tsx

- *Discovery*: Found original implementation in a1b2c3d.

  1. Context Recovery: Locate track-45 in conductor/archive/.
  2. Outcome: Re-implementation possible with original context.

📦 Git Internal Object Analysis

Git stores everything as "objects".

Inspecting a Blob

git cat-file -t <hash>  # Type check
git cat-file -p <hash>  # Content check

Finding Lost Blobs

git fsck --lost-found

💬 Forensic Interviewing of Code

  • "Who authorized you?": Authored-by, PR links.
  • "What was your purpose?": Conductor specs.
  • "Why were you killed?": Surroundings of deletion.
  • "Who were your friends?": git show --name-only.

🎙 Advanced Forensic Interviewing Scenarios

Scenario 1: The "Legacy Debt" Interview

When you find a block of commented-out code that says // DO NOT REMOVE - CRITICAL FOR IE11.

  • Question: Is IE11 still supported in the 2026 build?
  • Action: Check conductor/product-guidelines.md for current browser support tiers.
  • Follow-up: If no, use git log -S "IE11" to find the commit where support was officially dropped.

Scenario 2: The "Silent Fix" Interview

When a bug suddenly disappears but no one knows why.

  • Question: Which commit fixed the issue without mentioning it in the message?
  • Action: Use git bisect between the last "broken" version and the first "fixed" version.
  • Outcome: You find a refactor that accidentally fixed a race condition by moving a hook call.

🌍 Historical Contextualization

  • Temporal Correlation: Commits vs. Server Logs.
  • Environment Archeology: package.json / Dockerfile history.
  • Branch Lineage: Where did this change travel?

🤖 Integrating with AI Analysis

  1. Wide Sweep: Parallel grep and log.
  2. Filter & Rank: Candidate identification.
  3. Deep Read: Diff analysis.
  4. Synthesis: "Why" explanation.

📊 Comparison of Search Tools

ToolBest ForSpeedDeep History?
grep -rQuick scan.FastNo
ripgrep (rg)High-speed live search.Ultra FastNo
git log -SString lifecycle.SlowYes
git log -GRegex history.Very SlowYes

✅ Best Practices for Searchability

  • Atomic Commits: One change, one commit.
  • Semantic Messages: Prefixes and clear intent.
  • Reference Everything: Track IDs, PRs.
  • Tagging: Milestone markers.

🚫 The "Do Not" List (Common Pitfalls)

Anti-PatternWhy it failsCorrect Approach
Grep on live onlyMisses history.Use git log -S.
Ignoring ReflogMisses "undone" work.Check reflog first.
Searching without -pNo code visibility.Always use --patch.
Partial SearchMisses other branches.Use --all.

💡 Step-by-Step Implementation Patterns

Scenario: Restoring a deleted config

  1. Identify: git log --all --full-history -- config/settings.yaml
  2. Preview: git show <hash>:config/settings.yaml
  3. Recover: git checkout <hash>^ -- config/settings.yaml

🚀 Common Forensic Scenarios (2026 Edition)

Scenario A: The "Accidental Rebase" Recovery

You rebased your branch and lost work.

  • Action: git reflog. Find the hash before the rebase.
  • Fix: git reset --hard <safe_hash>.

Scenario B: The "Vanishing API Key"

Key accidentally committed.

  • Action: git log -G "AI_KEY_PATTERN" --all --patch.
  • Result: Immediate rotation of the found key.

🧠 Mental Models for Forensic Excellence

  • The Crime Scene Model: Document current state first.
  • The Tree of Life Model: Visualize the branch graph.
  • The Incremental Disclosure Model: Broad search -> Zoom in.

🛠 Advanced Troubleshooting & Recovery

Dealing with "Shallow" Clones

If the project was cloned with --depth 1, history search will fail.

  • Symptom: git log -S returns nothing even for known strings.
  • Fix: git fetch --unshallow to retrieve the full history.

Recovering from a Corrupt Index

  • Symptom: fatal: index file corrupt.
  • Fix: rm.git/index && git reset.

Finding "Dangling" Work

  • Action: git fsck --lost-found.
  • Outcome: Inspect .git/lost-found/commit/ for commits that aren't on any branch.

🔄 The Bisect Walkthrough: Automated Regressions

  1. Start: git bisect start
  2. Define Limits: git bisect bad HEAD; git bisect good v1.5.0
  3. Automate: git bisect run bun test
  4. Finalize: git bisect reset

🔄 Restoring Deleted Branches & History

# Find deleted branch hash
git reflog | grep "feature/name"
# Restore
git checkout -b feature/name <hash>

📁 Searching for File Renames & Moves

git log --follow -- <new_path>

🛡 Archival Integrity & Verification

  • GPG Signing: git log --show-signature.
  • Hashes: Verify metadata JSONs in Conductor.

⚖️ Legal & Compliance Search (Audit Trails)

  • Author Activity: git log --author="Name".
  • Audit Export: git log --pretty=format:"%h %ad %s" > audit.txt.

🔐 Archival Security & Privacy

  • GDPR: History contains personal data.
  • Secret Hygiene: Old keys are still dangerous.

🤖 Search Scenarios for 2026 AI Teams

  • Refactor Audit: AI-driven breaking changes.
  • Agent Mission Reconstruction: Auditing autonomous agents.
  • Prompt Injection Discovery: Searching malicious prompt changes.

🌐 Integrating with External Search APIs

  • GitHub API: Search across many repos.
  • Sourcegraph: Cross-repo forensics.

🛠 Recovery from Corrupt Repositories

  1. Fix Index: rm.git/index && git reset.
  2. Recover Objects: git fsck.

🔮 Future-Proofing Your Archives

  • Standard Formats: Markdown/JSON.
  • Avoid Dead Links: Embed context in commits.

❓ Common Questions and Answers (FAQ)

Q: Can I search for a string that was in a file I renamed? A: Yes. Use git log --follow -S "string" -- path/to/new_name.ts.

Q: How do I find commits by a specific person that touched a specific directory? A: git log --author="Name" -- path/to/directory/.

Q: I ran 'git gc' and now my reflog is smaller. Can I get it back? A: No. git gc prunes old entries.

Q: What if the commit message is just a Jira ID? A: Use a browser tool to fetch the context or check Conductor archives.


📖 Glossary of Forensic Terms

  • Pickaxe: String count lifecycle search.
  • Reflog: Black box of branch head moves.
  • Blob: File content object.
  • Tree: Directory structure object.
  • Dangling Commit: Unreachable commit.
  • Bisect: Binary search regression finder.
  • OID: Object ID (Hash).
  • Plumbing: Low-level Git commands.
  • Porcelain: High-level Git commands.
  • Detached HEAD: Working on a commit instead of a branch.
  • Staging: The prep area for commits.
  • G-Regex: The use of git log -G for pattern matching.
  • Diff-Filter: Filtering by addition (A), deletion (D), or modification (M).
  • Full History: Disabling parent simplification for total visibility.
  • Cherry-Pick: Applying a single commit from history to your current branch.
  • Revert: Creating a new commit that inverses the changes of a historical commit.
  • Unshallow: Converting a partial history clone into a complete one.
  • FSCK: File System Check for repository integrity.
  • Commit Graph: The directed acyclic graph (DAG) of all commits.
  • Reflog Expiration: The time-to-live of reflog entries (defaults to 90 days).
  • Blob Hash: The unique SHA identifier for a specific version of a file.
  • Git Flow: A branching model for release management.
  • Trunk-Based Development: A development model where developers frequently merge small updates to a core "trunk".
  • Merge Conflict: A state where Git cannot automatically resolve differences between two commits.
  • Remote: A version of the repository hosted on the internet or a network.

📜 Tooling & Scripts

  • deep-search.sh: Multi-branch search wrapper.
  • track-forensics.py: Squaads mission re-constructor.

📚 Reference Documentation


*Updated: January 22, 2026 - 15:18*

平台分布

Codex

32.63%
按下载量换算39

Claude

33.19%
按下载量换算39

Cursor

19.2%
按下载量换算23

Gemini CLI

8.47%
按下载量换算10

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills