GitHub Trending Daily — 2026-05-29
KST 2026-05-29 refresh. The chart shifted back toward practical transformation pipelines at the top, while the agent-quality tail stayed dense with taste, harness, and skill-packaging tooling.
Snapshot
Top 10 by daily star velocity:
| # | Repo | ★ today | Lang | Role |
|---|---|---|---|---|
| 1 | harry0703/MoneyPrinterTurbo | 3,563 | Python | AI video generation |
| 2 | microsoft/markitdown | 1,876 | Python | document-to-Markdown |
| 3 | EveryInc/compound-engineering-plugin | 354 | TypeScript | agent setup/plugin layer |
| 4 | twentyhq/twenty | 575 | TypeScript | AI-ready CRM |
| 5 | anthropics/claude-code | 460 | Python | terminal coding agent |
| 6 | Leonxlnx/taste-skill | 2,066 | Shell | taste guardrail |
| 7 | cursor/plugins | 129 | TypeScript | plugin spec / ecosystem |
| 8 | run-llama/liteparse | 680 | Rust | document parsing |
| 9 | galilai-group/stable-worldmodel | 346 | Python | world-model research |
| 10 | byoungd/English-level-up-tips | 1,564 | — | learning guide |
What changed in this refresh
MoneyPrinterTurboandmarkitdownnow sit at the top of the chart, which pushes the most visible signal toward media generation and document transformation.anthropics/claude-code,cursor/plugins,run-llama/liteparse, andDataTalksClub/data-engineering-zoomcampjoined the visible set, widening the chart beyond pure agent tooling.- The quality-control cluster remains dense:
taste-skill,stop-slop,ECC, andcompound-engineering-pluginstill point to the same demand for better output shaping and reuse.
Interpretation
This refresh suggests GitHub Trending is still rewarding AI-adjacent work, but the center of gravity is more balanced than a pure agent-framework chart. The strongest projects are either turning messy input into structured output or turning model output into something more controllable and reusable.
That is why MoneyPrinterTurbo and markitdown matter together: one converts prompts into short videos, the other turns files and office docs into Markdown. They are different domains, but they share the same winning pattern — reduce friction between raw material and something the next step can consume immediately.
The agent stack is still visible underneath that layer. taste-skill, stop-slop, ECC, anthropics/claude-code, and cursor/plugins show that the market still wants a better way to package behavior, not just a better model.
See also: 2026-05-28-github-trending-daily, moc-dev-tools, moc-ai-coding, moc-ai-agents.