BJ

battlewen/jailbreakskill

Developer tools

このページは AI により自動翻訳されています。技術的な詳細は原文をご確認ください。 原文を見る

27 stars 品質 62 トレンド 19

Code for paper "JailbreakSkill: Scaling Automated Red-Teaming with Reusable and Ever-Evolving Skills."

概要

Scaling Automated Red-Teaming with Reusable and Ever-Evolving Skills JailbreakSkill is a two-stage LLM red-teaming framework built around reusable prompt-rewrite skills. Stage 1 searches a fixed skill library with shared UCB memory; Stage 2 analyzes failures and evolves category-specific skills. The repository also includes tools that extract new skills from external evidence. This repository contains adversarial prompts and red-teaming transformations. Use it only on systems you own or are authorized to evaluate, with appropriate access controls, monitoring, and human review. - — 🎉 JailbreakSkill was released for the first time. - — 🛠️ We are developing a new skill for prompt injection. - — 🧩 If time permits, we plan to extend the pipeline with an agent runtime and a ReAct-style interaction loop, eventually enabling red-teaming of agentic targets. This will require corresponding updates to the harness.

README

以下は原文 README の抜粋です。原文を見る。

Scaling Automated Red-Teaming with Reusable and Ever-Evolving Skills

JailbreakSkill is a two-stage LLM red-teaming framework built around reusable prompt-rewrite skills. Stage 1 searches a fixed skill library with shared UCB memory; Stage 2 analyzes failures and evolves category-specific skills. The repository also includes tools that extract new skills from external evidence.

[!WARNING] This repository contains adversarial prompts and red-teaming transformations. Use it only on systems you own or are authorized to evaluate, with appropriate access controls, monitoring, and human review.

🔥 News

  • 2026-08-18 — 🎉 JailbreakSkill was released for the first time.
  • In progress — 🛠️ We are developing a new skill for prompt injection.
  • Planned — 🧩 If time permits, we plan to extend the pipeline with an agent runtime and a ReAct-style interaction loop, eventually enabling red-teaming of agentic targets. This will require corresponding updates to the harness.

💡 Motivation

From external attack knowledge to reusable, diagnosable, and ever-evolving jailbreak skills. Click the figure to view the original PDF.

🔄 Pipeline

Seed Prompts
     │
     ▼
┌─────────────────────────────────────────────┐
│  Stage 1 — UCB Skill Search                 │
│                                             │
│  UCB Memory ──rank──► Rewrite Skills (×16)  │
│      ▲                       │              │
│      │ score           rewritten prompt     │
│      │                       ▼              │
│   Judge  ◄──response── Target Model         │
└─────────────────────────────────────────────┘
     │                    │
   success             failure
     │                    │ (grouped by risk category)
     │                    ▼
     │   ┌─────────────────────────────────────┐
     │   │  Stage 2 — Skill Evolution          │
     │   │                                     │
     │   │  Failure Analyzer                   │
     │   │       │ dispatch                    │
     │   │       ▼                             │
     │   │Meta-Skills (refine/combine/discover)│
     │   │       │ new skill                   │
GitHub で README 全文を見る

推奨ツール

別のキーワードを試すか、フィルタを外してください。

インストール

npx skillfish add battlewen/jailbreakskill
MCP 初心者?mcpskill.org のガイドから始めましょう →