togg53192-cmd/jailbreaks
LIST OF ALL MY JAILBREAKS
★ 13,335 · Fork 1,990 · 在 GitHub 查看
在手机上浏览源码、订阅每日 AI 趋势日报与项目架构深度解读:获取 CopoHub 客户端(iOS & HarmonyOS)→
GitHub Trending 上榜记录
togg53192-cmd/jailbreaks 已在 GitHub Trending 每日榜单上榜 7 天,最高排名第 10,首次上榜 2026-10-27,最近一次上榜 2026-11-02。
| 日期 | 排名 | Star | 当日增长 |
|---|---|---|---|
| 2026-11-02 | #13 | 13,335 | +726 |
| 2026-11-01 | #14 | 12,907 | +1,924 |
| 2026-10-31 | #13 | 11,768 | +2,055 |
| 2026-10-30 | #10 | 10,887 | +2,053 |
| 2026-10-29 | #10 | 8,905 | +1,610 |
| 2026-10-28 | #10 | 7,027 | +1,478 |
| 2026-10-27 | #10 | 5,445 | +890 |
同期上榜的项目
- vercel-labs/agent-desktop – Computer use CLI for AI agents
- thedotmack/claude-mem – Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
- D4Vinci/Scrapling – 🕷️ The BEST Web Scraping framework that handles everything from a single request to a full-scale crawl! Don't be shy, join here: https://discord.gg/EMgGbDceNQ and follow here for daily tips and tricks: https://x.com/Scrapling_dev
- hyprwm/Hyprland – Hyprland is an independent, highly customizable, dynamic tiling Wayland compositor that doesn't sacrifice on its looks.
- Donchitos/Claude-Code-Game-Studios – Turn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.
- FlashML-org/FreeToken – FreeToken brings datacenter-scale model serving to your desktop. Run massive models locally, fast and efficiently.
togg53192-cmd/jailbreaks 项目深度剖析与商业评估报告
togg53192-cmd/jailbreaks 项目深度剖析与商业评估报告
分析前置声明:本仓库本质是一个纯文本资产库(Prompt/配置资产集合),而非传统意义上的软件工程项目。它没有可执行代码、没有构建产物、没有依赖树。因此本报告将把"架构"重新定义为提示词工程(Prompt Engineering)的资产组织架构,把"商业价值"重新定义为LLM 安全与红队领域的知识资产价值,并严格区分"技术深度"与"内容深度"。这一视角转换是理解该仓库全部价值的前提。
---
1. 核心定位与商业落地价值
1.1 根本痛点:指令层的"考古学"与不可复现性
该仓库 README 结尾一句"Star the repository if the collection saved you a session of prompt archaeology"精准点出了其核心痛点。前沿模型(Claude、Gemini、Grok、GLM、Kimi、DeepSeek 等)的系统提示词与指令层级(instruction hierarchy)是闭源、高频轮换、版本强耦合的。安全研究者、红队工程师与对齐研究者面临三重困境:
1. 不可复现:模型一次 OTA 更新即可让昨日有效的 persona lock 失效,缺乏版本化的实验基线; 2. 碎片化:同一技术(如"人格锁定 + 操作底线声明")在不同模型上需要不同的表述范式,散落在论坛、Twitter、Discord 中; 3. 格式异构:system message、AGENTS.md、Ollama Modelfile、Grok <CFG> 配置块——同一意图需要适配四种截然不同的注入面(instruction surface)。
该仓库以 31 个完整提示词 × 25+ 目标模型,将上述碎片整理为"paste-ready"的可复现资产,本质上是在为 LLM 安全研究建立一套最小可用的语料基线(corpus baseline)。
在 README 中展示上榜徽章
[](https://github.fq6825.top/repo/togg53192-cmd/jailbreaks)