Summary

This repository offers a specialized "reverse-flow skill" designed for AI agents and Codex, automating local Capture The Flag (CTF) reverse engineering workflows. It provides a structured process for analyzing binaries, firmware, and other artifacts within sandboxed environments, guiding users through stages from initial analysis to vulnerability assessment and reporting, thereby accelerating the reverse engineering lifecycle for security researchers and CTF players.

How It Works

The skill implements a deterministic workflow: Analyze → Report → Reverse → Deep Reverse → Vulnerability Assessment → User Choice. It leverages a hybrid prompt strategy, using English for internal execution rules, tool selection, and process control to enhance model stability, while defaulting to Chinese for user interaction, report structures, and next-step menus. This approach normalizes colloquial user requests (e.g., "unlock X," "remove X," "bypass anti-debug") into specific reverse engineering actions, facilitating smoother AI-driven analysis.

Quick Start & Requirements

Installation involves copying the skills/reverse-flow directory into the Codex skills directory (e.g., ~/.codex/skills/reverse-flow). The project includes example Python scripts (create_case.py, triage_artifact.py, tool_audit.py, report_from_triage.py) for automating case setup, artifact triage, tool auditing, and report generation within a specified working directory. No specific non-default prerequisites beyond the Codex environment are explicitly listed.

Highlighted Details

• Hybrid prompt strategy: English kernel prompts for stability, Chinese user interaction for accessibility.
• Structured workflow: Automates a multi-stage reverse engineering process.
• Colloquial request normalization: Translates natural language requests into actionable reverse engineering tasks.
• Default context: Assumes operation within local CTF, crackme, wargame, or training environments, reducing repetitive user input.

Maintenance & Community

No specific details regarding contributors, sponsorships, community channels (like Discord/Slack), or roadmap are provided in the README.

Licensing & Compatibility

The project is released under the MIT License. This permissive license generally allows for commercial use, modification, and distribution without significant restrictions, making it compatible with closed-source projects.

Limitations & Caveats

The skill is primarily designed for local, sandboxed, or offline analysis environments. Its effectiveness is dependent on the capabilities of the underlying AI Agent/Codex framework and the quality of the prompts used. No specific performance benchmarks or unsupported platforms are detailed.