REA: understand a feature, build your own version

Get your first useful result

Start with one feature in a JavaScript or Electron app you can inspect. Static analysis does not need Hopper, Ghidra or IDA. Have Node.js 22.19+ (22.x), 24.11+ (24.x) or 26+ and npm ready.

1. Connect your agent, then restart it

   Start with one feature in a JavaScript or Electron app you are entitled to inspect. Static JavaScript analysis does not require a native analysis engine. Check the current upstream installation guide for supported Node.js versions.

   npx rea-agents@latest setup

2. Copy this prompt and choose one feature

   Replace [APP PATH] and [FEATURE]. Try search, clipboard formatting or a keyboard shortcut. Keep the first task to analysis and a small build plan.

   Use REA to inspect [APP PATH] and explain how [FEATURE] works. Start with a summary, then trace only the relevant modules or functions. Cite files, symbols or addresses for each finding, and separate observations from assumptions. Return a short explanation, the evidence, anything still unknown, and the smallest implementation plan for my project.

3. Check the evidence before building

   A useful first result contains a clear explanation, source locations and a small implementation plan. Open one cited location and verify it. Then ask your agent to implement just that feature and compare its behavior on the same input.

Official READMESkill entry and workflowInstallation, clients and dependencies

What people are making with REA

Two first-person examples from X, with the original posts embedded below. These are creator-reported results; use each feature as a starting point for your own experiment.

Watch the REA walkthrough

Eric Tech introduces REA’s Skill and MCP tools, walks through setup with Hopper, and demonstrates a Wispr Flow-style app with a hotkey, note taking and local transcription.

This walkthrough is in English. Setup starts at 02:06, and the app demo starts at 05:01.

Eric Tech: How To Use REA To Clone Any Software In Minutes (YouTube)

Об этом Skill

REA connects a coding agent with tools for investigating shipped applications and their behavior. It routes different targets, such as native binaries, packaged apps, Electron and JavaScript, to suitable analysis workflows. Use the evidence to answer a specific question about a feature; for an ordinary source repository, the current instructions recommend normal source-analysis tools.

Почему мы его выбрали

Start with a small feature you can verify: a search interaction, clipboard behavior or a hotkey. REA supplies the analysis tools; your agent explains the findings and helps implement and test your version.

Для каких задач подходит

• Developers investigating how a specific feature works in a shipped application.
• Analysts comparing application builds, inspecting packaged resources or tracing binary functions.
• Teams using evidence from their own software to plan a reconstruction or compatibility implementation.

Choose the workflow by target

Official READMESkill entry and workflow

Install instructions, then connect REA tools

The skills CLI installs reverse-engineer-anything and its bundled references. It does not register the MCP service, install an analysis engine or add tools to an already running agent.

For the complete integration, run REA’s setup, select the intended clients, review its changes, then restart the agent. Setup installs matching workflow instructions by default. If tools already work, use them directly; run scoped diagnostics only for a missing connection or provider.

The supported agents shown here are the intersection of SkillHey’s shared catalog and upstream documented setup/manual integration clients. Other MCP-capable clients may use upstream manual registration.

Skill entry and workflowInstallation, clients and dependencies

From a question to an evidence-based explanation

1. Supply a specific target and question

   Give a local app, binary, ASAR, package or browser target and specify the feature or behavior to understand.

2. Inspect a useful overview and follow relevant paths

   Choose the target-appropriate tool; use a summary first for large artifacts, then follow selected functions, modules, imports or IPC boundaries.

3. Preserve evidence and unknowns

   Keep Evidence IDs and limitations with each finding. Request smaller retained views or an evidence export when a result is too large.

4. Explain or reconstruct the requested feature

   Use the findings to explain behavior or guide a separate implementation and its tests, with observed behavior distinguished from design choices.

Official READMESkill entry and workflowArtifact and JavaScript reconstruction guide

Outputs, release compatibility and limits

Native analysis returns pseudocode and assembly, while JavaScript analysis returns recovered modules and relationships. These are analysis results; exact recovery of the original source code is not guaranteed.

Static inspection reads supplied artifacts without executing the application. Runtime workflows can start or interact with the selected target. REA analyzes locally; model-provider processing of tool results follows your agent service’s terms.

Repository-main instructions may describe capabilities not yet present in an older npm release. Select features using the connected server’s actual tool list and input schemas, and refresh registrations and restart the client when updating.

Use targets you are entitled to inspect. Native format and host support vary by provider; consult the matching upstream guide rather than assuming every workflow is portable.

Official READMESkill entry and workflowInstallation, clients and dependenciesMIT license

Official README images and source review

The gallery uses the author’s official Hopper screenshot, investigation-flow diagram and case-study artwork. Open the source links for the original context.

MIT licenseAuthor’s Hopper screenshotAuthor’s investigation-flow diagramAuthor’s case-study artwork

When to use REA and what to prepare

Use REA when the answer depends on a shipped artifact, a supported analysis provider or a requested runtime observation. For a source repository you can already inspect, start with ordinary code search and tracing.

Installing Skill instructions alone does not install analysis engines or automatically configure every client. The upstream documentation describes plugin registration as an exception; check the actual host and available tools. Treat pseudocode and recovered modules as evidence with limits, not guaranteed original source.

Installation, clients and dependenciesReviewed upstream Skill

Перед использованием

• A local agent with REA MCP tools available, or a shell workflow using the REA CLI. The Skill alone supplies instructions and references.
• Node.js 22.x (22.19+), 24.x (24.11+) or 26+, plus npm. See the upstream setup guide for supported clients and host prerequisites.
• Deep native analysis needs a supported Hopper, Ghidra or IDA installation. Static JavaScript/Electron and .NET inspection do not require a native analysis engine.
• Extra targets have their own dependencies, such as a Chrome-family browser for website inspection, JADX and a JDK for Android code, or Binwalk/Unblob on Linux for firmware.
• REA is MIT-licensed. AI services and separate analysis tools may have their own costs; Hopper and IDA licensing are independent of this Skill.

Поддерживаемые агенты