Reverse-Skill Architecture Explained: 7 Key Components of the Modular Security Framework

The reverse-skill architecture consists of three orchestrated layers—Routing, Execution, and Output—supported by a self-installing bootstrap system and automatic evolution mechanism.

The reverse-skill repository implements a modular, self-evolving workflow engine designed for security-related reverse-engineering and penetration-testing tasks. Understanding the reverse-skill architecture is essential for security analysts who need to extend the framework or debug complex tool chains. This article breaks down the seven core components as documented in docs/ARCHITECTURE.md.

Routing Layer: SKILL.md and routing.md

The Routing Layer serves as the central nervous system of the reverse-skill architecture. Two files control all dispatch decisions:

• SKILL.md – The master entry point that declares all top-level skills available to the engine.
• routing.md – A keyword-to-sub-skill matrix that determines which module handles a given request.

When a user submits a security or reverse-engineering request, the engine first scans for trigger keywords. If a match exists in the routing matrix, execution proceeds; otherwise, the system suggests adding a new skill to the framework. This design is documented in ARCHITECTURE.md lines 7–14.

Execution Layer: Five Module Categories

The Execution Layer contains specialized sub-skills organized into five functional categories. According to the "Skills 模块关系图" section (lines 38–90), these include:

Reverse-Engineering Modules

• apk-reverse – Android package analysis
• ida-reverse – IDA Pro integration workflows
• radare2 – Open-source binary analysis
• binary-diff – Comparative binary analysis
• patch-diff-exploit – Patch-based exploit development

Exploitation Modules

• pwn-chain – Return-oriented programming chains
• firmware-pentest – Embedded device testing
• edr-bypass-re – Endpoint detection response evasion

Penetration-Test Modules

• pentest-tools – General offensive security utilities
• src-hunter – Source code discovery and analysis

Web/Browser Modules

• js-reverse – JavaScript deobfuscation and analysis
• browser-automation – Scripted browser interaction

Each sub-skill runs its own workflow and reports completion status to the central journal, as noted in lines 20–31.

Tool Discovery and Bootstrap System

Before any sub-skill executes, the framework verifies dependencies through tool-index.md. Missing tools trigger the bootstrap-reverse script system:

Platform	Script	Path
Windows	bootstrap-reverse.ps1	skills/scripts/bootstrap-reverse.ps1
Kali Linux	bootstrap-reverse.sh	kali/scripts/bootstrap-reverse.sh

The bootstrap process reads bootstrap-manifest.json and selects an installation method from five supported types: github-release-zip, pip-package, npm-global, winget-package, or local-http-mcp. After installation, binaries are added to PATH and tool-index is refreshed. This flow is detailed in the "Bootstrap 自举流程" section (lines 107–127).

Output Layer: Reporting and Visualization

The Output Layer transforms execution results into actionable intelligence through three components:

• docs-generator – Creates structured HTML reports from task metadata
• diagram-generator – Produces visual workflow charts
• field-journal – Persistent log storage that feeds the evolution engine

This layer executes after every completed task, as described in lines 32–40.

Automatic Evolution Mechanism

The reverse-skill architecture implements self-improvement through feedback loops. Every completed task writes a concise entry to field-journal/, which triggers cascading updates to:

1. _index.md – Internal skill index
2. routing.md – Routing matrix refinements
3. tool-index.md – Tool status refresh

This "自动进化机制" (lines 152–164) ensures the framework continuously learns from operational experience without manual intervention.

Bootstrap Commands and API Usage

The following commands demonstrate practical interaction with the reverse-skill architecture—each invoking specific components documented in the source:

# Bootstrap missing tools on Windows

.\skills\scripts\bootstrap-reverse.ps1 -Force

# Bootstrap missing tools on Kali Linux

bash kali/scripts/bootstrap-reverse.sh

# Dispatch a specific sub-skill via the routing layer

.\skills\scripts\master-route.ps1 -Hint "apk-reverse"

# Generate final report through the output layer

python docs/generator.py --output report.html

The master-route.ps1 and master-route.sh scripts implement the core routing logic, reading SKILL.md and routing.md to dispatch workflows.

Core Files Reference

File	Purpose	Location
SKILL.md	Master skill declaration	skills/SKILL.md
routing.md	Keyword-to-skill routing matrix	skills/routing.md
bootstrap-reverse.ps1 / .sh	Cross-platform dependency installation	skills/scripts/, kali/scripts/
tool-index.md	Runtime-generated tool availability status	(generated)
ARCHITECTURE.md	Complete architectural documentation	docs/ARCHITECTURE.md
field-journal/	Persistent task logs and evolution data	field-journal/

Summary

• Three-layer design: Routing (dispatch), Execution (specialized modules), and Output (reporting)
• Self-installing bootstrap: Automatically provisions tools on Windows or Kali Linux via manifest-driven scripts
• Keyword-driven routing: SKILL.md and routing.md determine which sub-skill handles each request
• Five module categories: Reverse-engineering, exploitation, pen-testing, web/browser, and infrastructure
• Automatic evolution: field-journal entries continuously refine the routing matrix and tool index
• Cross-platform support: PowerShell and Bash implementations maintain parity across operating systems

Frequently Asked Questions

What is the entry point for adding a new skill to reverse-skill?

New skills require two modifications: add a declaration to SKILL.md for top-level visibility, then register keyword mappings in routing.md so the engine can dispatch requests appropriately. The routing scripts master-route.ps1 and master-route.sh read both files at runtime.

How does reverse-skill handle missing dependencies?

The framework checks tool-index.md before executing any sub-skill. Missing tools trigger bootstrap-reverse.ps1 or bootstrap-reverse.sh, which consult bootstrap-manifest.json to determine installation method and source. Successful installations update PATH and refresh the tool index automatically.

Can reverse-skill run on platforms other than Windows and Kali Linux?

The bootstrap system explicitly supports Windows (via PowerShell and winget) and Kali Linux (via Bash and standard package managers). Other distributions may work if they provide compatible package management, but the architecture documentation in ARCHITECTURE.md only guarantees functionality for these two platforms.

What drives the automatic evolution of the framework?

Every task completion writes to field-journal/, which triggers updates to three index files: _index.md for skill organization, routing.md for routing efficiency, and tool-index.md for accuracy. This feedback loop is implemented without requiring explicit user commands, as detailed in lines 152–164 of ARCHITECTURE.md.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}