Stars
自媒体视频发布前违规风险审核。检查画面/声音/文字/封面/评论/带货信息/资质/引流,输出 Pass→Blocker 五级风险报告。覆盖抖音/小红书/视频号/B站/快手+千川低质素材。Claude Code、codex、workbuddy、hermes skill
A Claude Code plugin that makes Claude Code the foreman: it plans and reviews, and hands coding work to Codex, opencode, pi and other CLI agents.
NodeCloak for Windows and macOS: isolated browser profiles, per-profile proxies, environment checks, and Claude / Codex / Gemini terminal launchers.
Official implementation of our paper "CoACT: Action-Preserving Observation Compression for Coding Agents".
🔥🔥 The extremely fast inference engine for decision model
In-page prompt injection artifacts
Requirement-driven evaluation harness for AI agents and LLM applications. Generate behavior-specific test cases, run them against any target (hosted models, callable wrappers, OTel-traced agents), …
A post exploitation framework / C2 for advanced Red Team operations
Curated, simulated dataset of security logs (Windows, Web, Firewall, Cloud) with embedded attack vectors for SOC analysts to practice SIEM ingestion, threat hunting, and log analysis.
SOTA model-agnostic quantization for GGUF, local inference and beyond. Quantize and run text, image, video and audio models with llama.cpp, vLLM, Diffusers and PyTorch.
🏝️ Code for our paper: "Privasis: Synthesizing the Largest 'Public' Private Dataset from Scratch"
Stealthy, Linux-first multi-operator C2 with a self-healing mesh, in-memory execution, AI-driven MCP assistant, and autonomous evasion loop.
A Hub/research area, for jailbreaking LLM's, and breaking their safety filters.
Open-source AI security research lab for reproducible adversarial testing, jailbreak resistance evaluation, and LLM guardrail validation across multiple model providers.
The ultimate cyber skill for an AI agent. One folder. One registered skill. A full red-team and blue-team arsenal behind it: offense, defense, forensics, detection, and the tools that build both.
GreyZone: Production-quality Agentic Application Threats
FraudBench is a benchmark for evaluating large language models' propensity to recommend, assist, or directly commit legal fraud
针对deepseekV4家族的破甲块有效兼容层,实测100+调用全绿
Official implementation of EvoHarmBench (EMNLP 2026)
AI 自主渗透测试 · 红队作战平台 | 多 Agent 自主推进 · 漏洞情报运营 · 后渗透与 C2 | LLM 三通道轮询 | 全程证据留痕 | 原百度攻防赛冠军项目独立演进版
CHATDIFY: End-to-End Dify Workflow Synthesis From Natural Language via Multi-Agent Collaboration.
一个 AI 攻防知识库。A knowledge base for AI security.
A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories
From Memory to Behavior: A Behavior-Aware Role-Playing Framework for Social Media Influencers
The Ultimate Open-Source Security Arsenal for Hackers, Enterprises, and AI Agents——面向极客、企业与 AI 智能体的全域开源网络安全工具矩阵
Open-source cybersecurity harness to build and deploy sovereign, persistent AI security agents (sentinels) that continuously find, prove and fix security risks. Web UI, CLI, CI and MCP.