system_prompts_leaks

Thanks to Open Source Agent Analytics Understand your agents in production System Prompts Leaks The full verbatim system prompts behind the most popular chatbots in the world. Carefully curated and complete. Feel free to open up a PR or get in contact if you got a different result. The Washington...

awesome-dfir-skills

🔥 awesome-dfir-skills 一个由社区驱动的 DFIR / 事件响应 技能 集合：可重复使用的提示词、工作流和辅助文件，帮助从业者更快行动、保持一致性，或许——还能多睡一会儿。 🚀 快速开始 1. 从 skills/README.md 中选择一个技能 2. 复制/粘贴到你的 AI 助手（Claude, Codex 等）中 3. 当提示时，将你的工件提供给它 4. 见证魔法发生 ✨ 专业提示: 保留像 timewindow 这样的占位符不变——当技能要求时再填入具体值。 🤔 什么是“技能”？ 将技能视为 IR 的作弊码 。它是一个小巧、可重复使用的工件，你可以复制/粘贴到 ...

dataset

🚀 CySecBench：基于生成式AI的网络安全专用提示数据集，用于基准测试大语言模型 🛡️ 规模最大、最全面的基于生成式AI的网络安全专用数据集，用于基准测试大语言模型 🌟 概述 CySecBench 论文提供了： 🎯 一个尖端的提示数据集 ，包含12662条针对网络安全挑战的提示。 🧠 新颖的越狱方法 ，利用提示混淆与改进技术。 📊 对大语言模型（如ChatGPT、Claude和Gemini）的全面性能评估 。 为什么选择 CySecBench？ 现有数据集过于宽泛，且往往缺乏对网络安全的专注。CySecBench 通过提供领域专用的提示...

CL4R1T4S

CL4R1T4S 面向所有人的人工智能系统透明度与可观测性！包含从 OpenAI、Google、Anthropic、xAI、Perplexity、Cursor、Windsurf、Devin、Manus、Replit 等提取的完整系统提示词、指南和工具——几乎所有主流 AI 模型与智能体！ 📌 为什么会有这个 "要信任输出，就必须理解输入。" AI 实验室通过大规模、不可见的提示词框架来塑造模型的行为。由于 AI 对越来越多的人来说是一个可信的外部智能层，这些隐藏的指令会影响公众的认知和行为。 这些提示词定义了： AI 不能说什么 它们被迫遵循哪些人设和功能 它们如何被要求撒谎、拒绝或转移...

promptmap

O O o.-. Humans, Do Not Resist! |/ ,-'-. / /, / /|.-.| / --O-- .--""" /\ /\ | o.o / Utku Sen's /|\ -'--' / /| | - | / | | | \ /| | | ' \ '/ \ ' | ' \ | ' / | ' \ | / | | | ./| /||| ./|||,| .// / / |-----| || || || || promptmap2 is a an automated prompt injection scanner for custom LLM...

redeval

RedEval - LLM 安全评估框架 一个全面的框架，用于通过系统性攻击和拒绝测试评估大型语言模型（LLM）的安全性。RedEval 提供了一个统一、安全且可扩展的平台，用于评估 LLM 对抗对抗性提示和有害内容的鲁棒性。它是论文 "RedBench: A Universal Dataset for Comprehensive Red Teaming of Large Language Models" 的一部分，该论文是一个用于全面红队测试 LLM 的通用数据集。 📦 数据集 ：RedBench 数据集在 HuggingFace 上公开可用，地址为...

burpgpt

!IMPORTANT Announcing the launch of BurpGPT Pro, the edition specifically tailored to meet the needs of professionals and cyber boutiques. Discover a host of powerful features and a user-friendly interface that enhances your capabilities and ensures an optimal user experience. To access these...

PraisonAI AgentOS - Information Disclosure

PraisonAI's AgentOS FastAPI application server exposes an unauthenticated GET /api/agents endpoint that lists every registered agent's name, role and the opening of its instructions system prompt. No authentication is enforced on the route, allowing a remote attacker to enumerate agent...

CVE-2026-107717

The Banks library (prior to version 2.5.0 ) is vulnerable to a prompt injection flaw in the Prompt.chat_messages() function. The issue stems from the software attempting to parse every line of rendered template output as ChatMessage JSON . Because ChatMessage.role accepts arbitrary strings, an at...

BrokenHill

!\ Broken Hill 标志 \https://assets.kitploit.com/production/public/readmes/44608/c14154675c6f33f5c3887887afcdb39351c82f269ee2772cf31847106a20a871.jpg Broken Hill Broken Hill 是一款已产品化、开箱即用的自动化攻击工具，它使用贪婪坐标梯度（greedy coordinate gradient, GCG）攻击生成精心构造的提示词，以绕过大型语言模型（LLM）中的限制。该攻击方法详见 Andy Zou、Zifan...

GHSA-3GH4-CGHQ-F8V4 Pydantic AI OpenTelemetry instrumentation: retry prompt content is not redacted when `include_content=False`

Summary Pydantic AI's OpenTelemetry instrumentation supports InstrumentationSettingsincludecontent=False to exclude message content — user prompts, model completions, tool call arguments and responses — from exported telemetry, so that agents can be monitored without sending sensitive content to...

CVE-2026-107293

Pydantic AI , a Python agent framework for Generative AI, is vulnerable to potential information disclosure in versions 0.3.4 through 1.107.4 and 2.27.1 . When OpenTelemetry instrumentation is configured with InstrumentationSettings(include_content=False), the framework may export retry prompts o...

CVE-2026-88257

Improper Input Validation vulnerability in BeamMCP.Schema in ScriptKittyOS beammcp allows an MCP client to reach a tool's dispatch function with arguments that violate the input schema the server advertised. BeamMCP.Schema.validate/2 checked type, required, additionalProperties, enum and numeric...

CVE-2026-88257

The beam_mcp library (versions 0.1.0 before 0.10.1 ) contains an Improper Input Validation vulnerability in BeamMCP.Schema . The validate/2 function only enforces constraints (type, required, enum, and numeric bounds) on top-level arguments. Consequently, nested objects and array items lack enfor...

EUVD-2026-94980

Improper Input Validation vulnerability in BeamMCP.Schema in ScriptKittyOS beammcp allows an MCP client to reach a tool's dispatch function with arguments that violate the input schema the server advertised. BeamMCP.Schema.validate/2 checked type, required, additionalProperties, enum and numeric...

CVE-2026-88257 beam_mcp: nested tool argument constraints advertised but not enforced

Improper Input Validation vulnerability in BeamMCP.Schema in ScriptKittyOS beammcp allows an MCP client to reach a tool's dispatch function with arguments that violate the input schema the server advertised. BeamMCP.Schema.validate/2 checked type, required, additionalProperties, enum and numeric...

React19-fix-vibecoders

React 19 & Next.js 安全修复提示（AI Agent 版） 🚨 严重安全漏洞 — CVE-2025-55182 与 CVE-2025-66478 本仓库包含可复用的提示词，帮助 AI 编码助手自动修复关键的 React/Next.js 安全漏洞。 📋 内容概览 SECURITYFIXPROMPT.md — 包含完整上下文的综合性详细提示 SECURITYFIXPROMPTSHORT.md — 简洁版本，便于快速分享 SECURITYPROMPTREADME.md — 详细使用指南 🚀 快速开始 选项 1：快速修复（推荐） 复制...

system_prompts_leaks — Updated!

Thanks to Open Source Agent Analytics Understand your agents in production System Prompts Leaks The full verbatim system prompts behind the most popular chatbots in the world. Carefully curated and complete. Feel free to open up a PR or get in contact if you got a different result. The Washington...

CVE-2026-105139 Obot 0.26.0 before 0.26.2 Authorization Bypass via vMCP Profile Prompts and Resources

Obot 0.26.0 before 0.26.2 contains an authorization bypass vulnerability that allows authenticated users matching any vMCP profile to reach prompts and resources of ungranted components. Because profiles were enforced only on tools, attackers can access prompts, resources, and resource templates...

CVE-2026-105139 Obot 0.26.0 before 0.26.2 Authorization Bypass via vMCP Profile Prompts and Resources

Obot 0.26.0 before 0.26.2 contains an authorization bypass vulnerability that allows authenticated users matching any vMCP profile to reach prompts and resources of ungranted components. Because profiles were enforced only on tools, attackers can access prompts, resources, and resource templates...