Most security reverse-engineering and pentesting tasks force an operator—human or AI—to choose among many specialized tools and ad-hoc playbooks. reverse-skill treats that decision layer as first-class: a Skill Router + Tool Orchestration system that turns task classification and triage into a deterministic workflow (classify → route → check tools/MCP → execute → report), reducing guesswork when agents handle APKs, binaries, frontend JS, firmware, or CTF challenges.
What Sets It Apart
- Skill Router: A routing matrix maps task type and intent to scenario skills (APK, ELF, JS, CTF, etc.), so an agent knows whether to use jadx, radare2, Frida, IDA, or BurpSuite instead of guessing.
- Tool Orchestration & Bootstrapping: Detects local tool status and exposes script entry points across platforms (Windows/Linux/macOS/Kali), enabling repeatable calls to real tools — this makes agent actions verifiable and reproducible.
- MCP / Execution Surfaces: Integrates MCP servers (Burp, IDA, browser analysis) and local CLI scripts so agents can reach remote execution surfaces and live debugging hooks, not just static analysis.
- Field Journal & Reporting: Captures timelines, evidence chains, findings and reusable lessons so post-task knowledge reduces repeated mistakes in future cases.
Who It's For and Tradeoffs
Great fit if you run code-first AI agents (Claude Code, Codex CLI, Cursor, Cline, etc.) and need them to perform structured reverse-engineering or pentesting tasks with real tools and reproducible playbooks. It helps teams that want audited, repeatable agent-driven workflows for APKs, binaries, JS obfuscation, firmware, and CTFs.
Look elsewhere if you want a turnkey SaaS vulnerability scanner or a single-tool installer: reverse-skill is a workflow OS that assumes access to local tools and operational knowledge (Java/JDK, Node.js, Python, Frida, IDA/Ghidra) and focuses on orchestration, not replacing those tools. It also requires attention to legal/authorization boundaries—scope gating is part of its RULES.
Where It Fits
This project complements existing reverse-engineering toolchains by adding an agent-facing layer that standardizes triage and execution. Use it to make AI agents less guessy and more auditable when they must call heterogeneous tooling across machines and environments.
