v2 (experimental) just shipped

greasebig

The Open Source toolkit for Computer Vision & AI research. Less slop, insights pop.

$ git clone https://github.com/greasebig
See daily papers Read the docs

Works with every agent.

One install. Daily papers, agent indexes, harness engineering — greasebig plugs into any workflow that supports open-source research tools.

📄
Daily Papers
Tracking Hub
🤖
Agent Index
Meta-index
⚡
Harness POC
Engineering
🔬
VideoCoF
CVPR 2026
🎯
LiveMoments
ICLR 2026
🎨
DynVFX
SIGGRAPH
🧠
MCP Research
Protocol
🌐
Open Source
Community
Show all 8 agents ↓

Escape
the generic
AI research.

Current skills.

Pick the right skill for the job.

Daily Video Papers
Tool
Automated arXiv paper tracking hub for Video, World Models, Agents & Tone/Color research. Daily updates with translated abstracts.
Install: daily-video-papers
Awesome Agent Everything
Resource
Meta-index of ALL awesome agent repos on GitHub — frameworks, skills, papers, MCP, harness & more. Auto-updated every 3 days.
Install: awesome-agent-everything
Harness POC
v2 experimental
Minimal Agent Harness POC — Agent = Model + Harness. ~300 lines of Python with multi-provider support, memory, permissions, and hooks.
Install: harness-poc
MCP Deep Dive
Visual style
全面解析 MCP 诞生、三大原语、规范演进、OpenAI 采纳分水岭、与 A2A 协议的垂直/水平互补关系及 2026 生态系统现状。
Read: mcp-deep-dive
Claw Runtime & Agent Ecosystem
Legacy
全面解析 OpenClaw 病毒式爆发、Agent Harness 工程化实践、MCP/A2A/ACP 三大协议生态，以及 AI Agent 技术演进路线图。
Read: claw-runtime
Harness Engineering
Visual style
从 Mitchell Hashimoto 的定义出发，深入探讨 Agent = Model + Harness 范式。分析 WorkBuddy 如何作为一套完整的 Harness 运行。
Read: harness-engineering
VideoCoF
CVPR
Unified video editing with temporal reasoner. State-of-the-art framework for coherent video manipulation.
Repo: VideoCoF
LiveMoments
ICLR
Key photo restoration in Live Photos via reference-guided diffusion. Reselect and restore your best moments.
Repo: LiveMoments
DynVFX
SIGGRAPH
Augmenting real videos with dynamic visual content. Seamlessly add effects to existing footage.
Repo: dynvfx
ReCo
CVPR
Region-constrained in-context generation for instructional video editing with precise spatial control.
Repo: ReCo
Foundation Model Challenge
Challenge
Reproduction and exploration of the CVPR 2023 Foundation Model Challenge, Track 2 — cross-modal track solution.
Repo: cvpr-2023-challenge
Building a Harness from Scratch
Tutorial
~300 lines of Python. Multi-provider support, Agent Loop, AGENTS.md context injection, MEMORY.md cross-session memory, permissions & hooks.
Read: harness-poc
Agent Framework Study
Image gen
Comprehensive study of agent frameworks, MCP protocol integration, and harness patterns for building reliable AI coding agents.
Explore: agent-study
Show all 13 ↓

What is new in

v2

A substantial 2026 rewrite of the default skill. The agent now reads the brief, infers the right direction, and ships interfaces that do not look templated. Same install command, much stricter ruleset.

§0
Brief inference
The agent reads the room before generating: industry, audience, mood, motion depth, layout family.
§2
Brief to design system map
When to reach for Material, Fluent, Carbon, Polaris, Atlassian, Primer, GOV.UK, USWDS, Bootstrap, Radix, shadcn, Tailwind official, versus native CSS.
§8
Dark mode protocol
Dual-mode by default. Contrast and hierarchy parity across themes.
§11
Redesign protocol
Audit-first on existing projects. Preservation rules and explicit modernisation levers.
§12
Block library schema
Contract for iterative block additions so the library stays coherent.
§14
Hard pre-flight check
Every checkbox must honestly pass before the agent ships output.
$ npx skills add Leonxlnx/taste-skill
Read full changelog CHANGELOG.md on GitHub
↗ 🏠

GitHub

Projects, articles & links

 ↗ 📚

Papers Hub

Daily arXiv paper tracker

 ↗ 🤖

Agent Index

Meta-index of agent repos

 ↗ ✍️

Articles

Tech deep dives & feeds

 ↗ 📊

Stats

Live numbers & metrics