GitHub Top 10, Week 38 of 2026
Sep 11, 2026 - Sep 17, 2026
The ten repositories that gained the most stars on GitHub in the week of Sep 11, 2026 - Sep 17, 2026, counted from GitHub's public event log. A star is a bookmark, not an endorsement: the commit counts and flags below show which ones people are actually building.
Partial data. GitHub's public event log is missing most events on Sep 14, 2026, Sep 17, 2026. Counts for this week are minimums, and a repo whose surge fell on those days may be ranked too low.
| # | Repository | New stars | Stars | Commits (30 days) | License |
|---|---|---|---|---|---|
| 1 | MrDiamond64/Hydra | 2,958 | 2,958 | 2 | GPL-3.0 |
| 2 | bilawalsidhu/gods-eye-view | 2,239 | 46,451 | 9 | NOASSERTION |
| 3 | ayghri/i-have-adhd | 1,922 | 52,882 | 1 | MIT |
| 4 | alibaba/open-code-review | 1,349 | 43,141 | 7 | Apache-2.0 |
| 5 | deepseek-ai/deepseek-harness | 1,345 | 242,195 | 0 | MIT |
| 6 | tt-a1i/archify | 1,292 | 76,165 | 8 | MIT |
| 7 | debpalash/VoiceStudio | 1,287 | 51,873 | 22 | AGPL-3.0 |
| 8 | DietrichGebert/ponytail | 1,076 | 151,758 | 1 | MIT |
| 9 | JustVugg/colibri | 1,045 | 39,031 | 9 | Apache-2.0 |
| 10 | hypit-ai/hypit | 988 | 18,715 | 6 | NOASSERTION |
1. MrDiamond64/Hydra
An Among Us utility and anticheat mod to enhance the Among Us playing experience
2,958 new stars this week. 2 commits and 4 closed issues in the 30 days to Sep 17, 2026. C#, GPL-3.0 license, created Feb 4, 2026.
2. bilawalsidhu/gods-eye-view
A spy satellite simulator in your browser, except the data is real. Live open source spatial intelligence on a photorealistic 3D globe.
2,239 new stars this week. 9 commits and 30 closed issues in the 30 days to Sep 17, 2026. JavaScript, NOASSERTION license, created Jun 22, 2026. Worth noting: no standard open-source license.
3. ayghri/i-have-adhd
A skill to stop your coding agent from burying the answer. ADHD-friendly output.
1,922 new stars this week. 1 commits and 1 closed issues in the 30 days to Sep 17, 2026. Python, MIT license, created May 13, 2026.
4. alibaba/open-code-review
Secure, fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level comments, built-in multi-language ruleset (NPE, thread-safety, XSS, SQL injection), OpenAI & Anthropic compatible.
1,349 new stars this week. 7 commits and 8 closed issues in the 30 days to Sep 17, 2026. Go, Apache-2.0 license, created May 18, 2026.
5. deepseek-ai/deepseek-harness
DeepSeek Harness: Everything is a Plugin.
1,345 new stars this week. 0 commits and 0 closed issues in the 30 days to Sep 17, 2026. TypeScript, MIT license, created Aug 13, 2026. Worth noting: no commits in the 30 days to the week's end.
6. tt-a1i/archify
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams - self-contained HTML with motion and crisp export.
1,292 new stars this week. 8 commits and 2 closed issues in the 30 days to Sep 17, 2026. JavaScript, MIT license, created Apr 15, 2026.
7. debpalash/VoiceStudio
VoiceStudio is the open-source, fully-local ElevenLabs alternative - voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.
1,287 new stars this week. 22 commits and 12 closed issues in the 30 days to Sep 17, 2026. Python, AGPL-3.0 license, created Apr 9, 2026.
8. DietrichGebert/ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
1,076 new stars this week. 1 commits and 0 closed issues in the 30 days to Sep 17, 2026. JavaScript, MIT license, created Jun 12, 2026.
9. JustVugg/colibri
Run frontier MoE models on hardware you already own - pure C, zero deps, experts streamed from disk. Tiny engine, immense model.
1,045 new stars this week. 9 commits and 9 closed issues in the 30 days to Sep 17, 2026. C, Apache-2.0 license, created Jul 1, 2026.
10. hypit-ai/hypit
Clone any viral video with AI agents. Not just a script, the whole workflow: swap the face, the words, the B-roll, ship 100 variants in one command, and get your 100M views.
988 new stars this week. 6 commits and 1 closed issues in the 30 days to Sep 17, 2026. TypeScript, NOASSERTION license, created Jul 29, 2026. Worth noting: no standard open-source license.
The week's story
Most of what rose on GitHub this week was not software. It was advice for agents, written in plain text.
Look at the top of this week's GitHub chart and count the repos that contain almost no code. There are a lot of them. A skill that makes your coding agent stop burying the answer. A skill that makes it think like the laziest senior developer in the room. Skills for architecture diagrams. Skills that strip the signs of AI from its writing. A security audit skill from Cloudflare.
A skill is a folder of instructions, mostly markdown, that a coding agent such as Claude Code or Codex loads and follows. No compiler. No runtime. Just text your agent treats as orders. And this week engineers starred them faster than most frameworks.
One note on the numbers before we start. They come from GitHub's public event log, which dropped a lot of events on September fourteenth and seventeenth. So every weekly star count here is a floor, not a total. Read the ranking and the shape, not the exact figures.
The pattern
Teams have a model. Most of them have an agent harness by now. What they lack is judgment in a form the agent can follow. So they share it the way engineers always shared judgment. As a style guide. Except now the reader is a machine, and it does exactly what the file says.
That makes these files powerful and a little dangerous. A skill runs with your agent's permissions. It can tell the agent to fetch things, run things, and write things. Treat it like a dependency, because that is what it is.
Now the picks.
Pick one. Open Code Review, from Alibaba.
It is a code review tool that pairs fixed, rule-based checks with a language model agent, and leaves comments on precise lines. It is rising because agent-written pull requests are piling up faster than people can read them. It gained at least 1,349 stars this week, and it has 192 contributors and 147 commits in the last thirty days, with regular releases. Apache licensed. Verdict: TRY. Who should care: any team where agents open more pull requests than humans review.
Pick two. Colibri.
It is a small engine, written in plain C with no dependencies, that runs large mixture-of-experts models on hardware you already own by streaming the experts from disk. At least 1,045 stars this week, 904 commits in thirty days, 193 contributors. That is a very busy young project. The claim that it runs frontier models on ordinary machines is the README's, not ours. Verdict: TRY, if you run models locally and have an afternoon.
Pick three. Orca, from Stably.
It is a desktop app for running a fleet of coding agents in parallel, each with your own subscription. At least 820 stars this week, over 2,300 commits in thirty days, near-daily releases, MIT licensed. The activity is real. Verdict: TRY. Who should care: developers already juggling three agent sessions in three terminals.
Pick four. Security Audit Skill, from Cloudflare.
It walks a coding agent through a security audit in phases and writes findings in a format a machine can check. At least 706 stars this week. It has four contributors and three commits in thirty days, which is normal for a file of instructions. Remember that Cloudflare sells security. The claim that findings are independently verified is the README's. Verdict: TRY, on a repo you know well, so you can judge what it misses.
Pick five. Matt Pocock's skills.
A public folder of the skills one well-known TypeScript teacher uses every day. At least 866 stars this week on top of 270,000. Eight contributors, MIT licensed. Verdict: TRY, but as reading. The value is seeing how a careful engineer writes instructions for a machine.
Pick six. WeKnora, from Tencent.
It turns raw documents into a searchable knowledge base with an agent on top and a wiki that keeps itself updated. At least 549 stars this week and 333 commits in thirty days. The license field says no assertion, so check the terms before you build on it. Verdict: WATCH.
Pick seven. ARTEMIS, from Google.
It turns plain-language instructions into Android automation, with logs, and plugs into coding agents. At least 951 stars in a week for a repo created in August, eight contributors, no release yet. Verdict: WATCH. Promising for mobile testing, too early to depend on.
Now the hype check.
The single most-starred repo of the week was not on our beat. It was a mod for the game Among Us, created in February. All 2,958 of its stars arrived inside this one week, and the event log shows one new fork. We make no claim about how those stars arrived. We only note that stars and use point in very different directions, which is why a star is a bookmark, not an endorsement.
The bench.
Ponytail, the lazy-senior-developer skill. Archify and Diagram Design, two skills for diagrams. Humanizer, which scrubs AI tells from writing. Superpowers, the skills framework that started much of this. Spec Kit from GitHub, for spec-driven development. OmniRoute, a free gateway to hundreds of model providers. VoiceStudio, a local voice cloning and dubbing studio. And God's Eye View, a browser globe fed by live public data.
Through-line: this week engineers starred instructions for agents faster than code for them.
One thing to try this weekend: pick one skill you are tempted to install and read every line before you do. If it tells your agent to fetch, run, or send anything you would not approve by hand, do not install it.
*How this is made: stars, forks, commits, issues and releases are counted from GH Archive, the public record of GitHub events, for the Friday-to-Thursday window. Lists, courses and prompt collections are left out. Descriptions are each project's own. *