チュートリアル · 投稿者 ZookeepergameMost817

AIリバースエンジニアリングツールキット：無料ツール、エージェント、コース、実践ラボ——理解したい対象別に整理

エージェント支援のリバースエンジニアリングの出発点として REA を紹介する Reddit のリソースマップ。セットアップ手順、実際の調査、ネイティブ、.NET、Electron、iOS、MCP ツールへのリンクを含む。

ZookeepergameMost817@ZookeepergameMost817

翻訳済み

この事例について

この投稿は無料のリバースエンジニアリングツールを対象別に整理し、REA を先頭に置いています。REA リポジトリ、インストールドキュメント、初回調査ガイド、Notes エクスポート、Notion クリップボード、DX-Ball、Aegis などの実践的なショーケースへのリンクがあります。npx rea-agents のセットアップコマンドを含み、商用エンジンとモデルプロバイダーのデータポリシーに関する注意を促したうえで、ネイティブバイナリ、.NET、iOS、Electron/JavaScript 向けの MCP ブリッジと分析ツールを列挙しています。これは単一のケーススタディではなく、教育リソースのまとめです。

セットアップ

インストール方法は元のソースをご確認ください。

原文 / README

You don’t need another folder of 150 bookmarks.

You need to know which tool to open when you’re staring at a Windows binary, an Electron app, a firmware image, or an undocumented protocol.

This is a reverse-engineering resource map organized by target—not a random list of repos.

AI agents are part of it. So are disassemblers, debuggers, protocol analyzers, and free courses that teach you how to check the agent’s work.

The rule that ties everything together: evidence.

A convincing explanation is not the same as a verified reconstruction.

01 — Start with REA: connect your agent to the investigation

REA connects coding agents to tools for inspecting native binaries, JavaScript/Electron applications, .NET assemblies, websites, and runtime behavior. Results include supporting evidence and limitations.

Start here:

Setup:

npx rea-agents@latest setup

Review the proposed changes, approve the setup, and restart your agent.

Two important caveats:

REA is open source; commercial analysis engines such as IDA Pro and Hopper still require their own licenses.
REA analyzes targets locally, but tool results go to your agent. Your model provider’s data policy still matters.

02 — Follow a real investigation before inventing your own workflow

Start with one feature, not “explain this entire application.”

Notes export — follow a button through the preload bridge and IPC handler to the CSV writer. Then predict an edge case and test it.
Notion clipboard — follow a clipboard API through Electron’s process boundaries.
DX-Ball — follow a native sound calculation from original instructions to a checked reconstruction.
Aegis — explore an APK investigation focused on TOTP behavior.

The Notes exercise is a particularly good first step: the app is small, the trace is concrete, and you can check the CSV formatter’s behavior directly.

Also worth exploring: CTF Flag Checker, REA Gems, Chrome Dinosaur, and the Notes Evidence Record.

03 — AI reverse-engineering agents and MCP bridges

For connecting an agent to an analysis environment:

Also explore: Revenant, Awesome LLM Reverse Engineering, AgentRE-Bench, SRE-Bench, and Reverse Engineering Agent Skill.

Think of the agent as an investigator’s assistant—not the final authority on what a binary does.

04 — Native binaries

The core toolkit:

Keep these nearby too:

Keystone · GNU Binutils · GDB · LLDB · pwntools · ROPgadget · Detect It Easy · capa · YARA

05 — .NET applications

Start here:

Also explore:

AsmResolver · Harmony · Mono.Cecil · NativeAOT reversing helpers

06 — iOS

Documentation, practice targets, and device tooling:

Also explore:

ios-deploy · class-dump · ipsw · Objection

07 — Electron and JavaScript

For package