Stars
Dual-engine (llama.cpp + vLLM) LLM benchmarking pipeline for GGUF & safetensors on NVIDIA GPUs — speed, quality, live dashboard, publishable cards.
A quantization compiler that delivers 12× lower KL than Q4_K_M at the same file size
Inkling 2 Mini on one RTX 3090: 128 tok/s, 344K context, images and tools. A Project Maya engine (maya-ik2) with every expert in VRAM.
97 tested hypotheses on making huge MoE LLMs fast on consumer GPUs (4x RTX 3090 + experts in RAM): what worked, what failed and why. A playbook for people and a ready skill for coding agents.
A Claude Code skill bundle for bug hunting and external red-team work - 82 skills, 15 slash commands, 681 disclosed-report patterns curated across 24 core vulnerability classes, plus enterprise ide…
A simple, lightweight PowerShell script that allows you to remove pre-installed apps, disable telemetry, as well as perform various other changes to declutter and customize your Windows experience.…
My agent skills and harness setup for Claude Code, Codex and T3 Code
Autonomous penetration-testing agent (AGPL-3.0). Authorized targets only.
Safely run your agents using any coding harness
Abliterates models quickly and efficently, Qwen2.5-7B: 95%→2% refusal at 0.095 KL
English and Traditional Chinese version of enhanced ARTEX
Reversible HTML and JavaScript prompt decoys with opaque per-insertion keys.
Multi-tier framework for evaluating AI agent skills and plugins with quality gates, security scanning, harness compatibility checks, semantic overlap detection, synthetic evaluation dataset generat…
A native control plane for Claude Code, Codex, Cursor, Devin and other coding agents.
One-menu RTX 5090 installer for fafstmobel with Cinference, DFlash2, K8V4, and vision.
MUCH Improved version of the Python Chasing Your Tail Tool to help you determine if you're being followed
Accelerate high-latency GEO inflight Wi-Fi via warm BBR TCP multiplexing + 1-RTT Ground-TLS + zlib (zero dependencies)
Open-source AI-powered offensive security harness for automated penetration testing.
An agent skill that keeps other skills lean and correct: a checklist for every edit, and a tested diet for bloated ones.
Qwen3.8-Flash-Next on 2x RTX 3090: 150 tok/s on code, 262K context. Strata 0.2.10 + unsloth UD-IQ4_XS + DFlash2. Full recipe, refusal vectors, session patch, raw benchmark traces.
📡 Your own AI-powered news radar. Generates daily briefings in English & Chinese. | 用 AI 构建你专属的新闻雷达
让你的命令行 Agent 在后台自己接活、自己跟进到底;平时不烦你,遇到拿不准的事才发卡片找你拍板。
Towards a More Autonomous Agent System
ARTEX - AI autonomous penetration testing system (English translation; upstream: Autumn-27/ARTEX, AGPL-3.0)
Easiest way to give your coding agents VMs - locally, remote, or cloud sandboxes
ReAlign models instantly, on demand. Per-request alignment for verified users and channels, or Always On. Use directly or with Hermes, OpenClaw and custom harnesses.
A Claude skill that turns vibecoded, AI-looking UI into a designed interface
A framework for few-shot evaluation of language models.