Reverse Engineering with Code Evaluation and the JEB Pro API

Analyze binaries using JEB Pro's Python API in headless mode. Use when examining program structure, functions, disassembly, cross-references, decompilation (native C / Dalvik Java / WASM), or strings without the GUI.

Provides:

• agent jeb-expert: Senior JEB Pro Python developer and reverse engineer. Use proactively when writing JEB scripts, debugging JEB API issues, analyzing binary analysis problems, or when the user needs expert guidance on reverse engineering tasks with JEB Pro.
• skill jeb-scripting: Write and execute Python scripts using the JEB Pro API for reverse engineering. Analyze binaries, extract functions, strings, cross-references, decompile code (native/Dalvik/WASM), work with JEB databases (.jdb2). Use when user wants to analyze binaries, reverse engineer executables, or automate JEB Pro tasks.

Installation

This plugin is local (not in a marketplace). From this directory:

# Claude Code (local plugin install)
claude plugin install $REPO_DIR

# Or symlink the skill into each CLI's discovery dir (cross-CLI):
# Codex:   ln -s $REPO_DIR/skills/jeb-scripting ~/.codex/skills/jeb-scripting
# zcode:   ln -s $REPO_DIR/skills/jeb-scripting ~/.agents/skills/jeb-scripting

First-time setup (per skill dir):

cd $REPO_DIR/skills/jeb-scripting && uv run python setup.py

Requirements

• uv package manager
• JEB Pro 5.0+ (Jython scripting engine)
• JDK 11+ for running JEB
• A valid JEB Pro license on this machine

Set JEB_DIR if setup can't auto-detect your JEB installation:

export JEB_DIR="/opt/jeb"               # Linux
export JEB_DIR="/Applications/JEB.app"  # macOS

Usage

Once the skill is loaded, ask the AI agent to analyze a binary:

"Analyze sample.exe — list all functions and their addresses"
"Find all strings containing URLs in app.apk"
"Decompile the main function in firmware.bin"

The agent will automatically write the appropriate JEB script and execute it.

Architecture

jeb-agent/
├── .claude-plugin/plugin.json    # Plugin registration
├── agents/jeb-expert.md          # Agent persona and expertise
├── commands/jeb-skill/
│   └── bootstrap.md              # API_REFERENCE.md generator
├── README.md
└── skills/jeb-scripting/
    ├── SKILL.md                  # Skill definition and usage patterns
    ├── API_REFERENCE.md          # JEB API quick reference
    ├── run.py                    # Universal script executor (headless wrapper)
    ├── setup.py                  # Environment setup and JEB detection
    ├── pyproject.toml            # Python project metadata (uv)
    └── .venv/                    # uv-managed virtual environment

How It Works

1. The AI agent writes a Python analysis script to /tmp/jeb-<timestamp>-<name>/script.py
2. run.py wraps the script as a JEB IScript implementation with ctx/prj context
3. JEB is launched in headless mode: jeb_macos.sh -c --script=<script> --infile=<binary>
4. Output is captured and displayed in real-time
5. Temp scripts are auto-cleaned