---
title: "這是我在嘗試逆向工程任何東西之前會使用的 REA 框架。給出"
url: "https://reamcp.dev/zh-tw/cases/this-is-the-rea-harness-i-d-use-before-trying-to-reverse-eng-414e9f"
source: "https://x.com/i/status/2110401001719321075"
source_type: "resource"
author: "Ricker"
author_handle: "0xRicker"
stars: 0
category: "檔案格式與協定"
categories: ["formats-protocols"]
tags: ["harness", "mcp", "rea", "reverse-engineering", "workflow"]
added: "2026-10-16T16:41:59.060Z"
updated: "2026-10-16T16:41:59.060Z"
locale: "zh-tw"
content_locale: "zh-tw"
---

# 這是我在嘗試逆向工程任何東西之前會使用的 REA 框架。給出

這是我在嘗試逆向工程任何東西之前會使用的 REA 框架。給出目標，告訴它你想理解什麼行為，它就會圍繞這個來構建調查，而不是盲目地翻找檔案。

## Description

這是我在嘗試逆向工程任何東西之前會使用的 REA 框架。

給出目標，告訴它你想理解什麼行為，它就會圍繞這個來構建調查，而不是盲目地翻找檔案。

它會找出路線，端到端地追蹤功能，將已確認的事實與猜測分開，並帶回你可以實際檢查的證據。

有用的部分是輸出：

• 系統在做什麼
• 行為在哪裡
• 哪些檔案 / 函式重要

比問 LLM「解釋這個應用」並得到一個沒有證據的自信故事要好得多。

提示詞 + 工作流程如下 ↓

## Original README

This is the REA harness I’d use before trying to reverse engineer anything.

give it the target, tell it what behavior you want to understand, and it builds the investigation around that instead of blindly digging through files.

it figures out the route, traces the feature end-to-end, keeps confirmed facts separate from guesses, and comes back with evidence you can actually inspect.

the useful part is the output:

• what the system is doing
• where the behavior lives
• which files / functions matter

much better than asking an LLM “explain this app” and getting a confident story with no proof.

prompt + workflow below ↓