Today's GitHub trending tells a story about infrastructure maturity: developers aren't just building with AI tools anymore — they're building around them. A cross-company Codex plugin, a 337-skill Claude Code library, and a fully local meeting assistant all point toward an ecosystem that's starting to define its own standards.
openai/codex-plugin-cc: OpenAI Builds a Plugin for Claude Code
OpenAI published a Claude Code plugin that lets you invoke Codex directly from within a Claude Code session — for code review or task delegation. With 1,532 stars in a single day, it's today's top mover. The existence of this plugin is itself the story: two competing AI labs now building interoperability tooling for each other's developer environments suggests that the developer tooling layer, not the model layer, is where the real integration work is happening.
github.com/openai/codex-plugin-cc
Zackriya-Solutions/meetily: AI Meeting Assistant That Never Leaves Your Machine
Meetily is a privacy-first meeting assistant built in Rust, offering live transcription through Parakeet/Whisper (4× faster than baseline), speaker diarization, and Ollama-powered summarization — all running locally with zero data leaving the device. It added 1,409 stars today. As enterprise hesitancy around sending meeting audio to third-party cloud APIs grows, fully offline AI meeting tooling is shifting from niche to necessary.
github.com/Zackriya-Solutions/meetily
alibaba/page-agent: Natural Language Control for Web Interfaces
Alibaba's page-agent embeds a GUI agent directly into web pages, letting users control web interfaces through plain-language instructions. The library picked up 805 stars today. It's part of a broader push toward in-browser agent primitives — closing the gap between describing what you want and actually doing it inside an existing web app, without rewriting the interface.
alirezarezvani/claude-skills: 337 Skills and Counting
A community-built collection of over 337 Claude Code skills, agent templates, and plugins — including 30+ pre-built agents and 70+ custom commands — sits at nearly 21,000 total stars and is among today's top Python repos. Alongside hesreallyhim/awesome-claude-code (nearly 49,000 stars), this catalog signals that Claude Code's skill system has crossed the threshold from experimental to something developers are actively building standards around.
github.com/alirezarezvani/claude-skills
usestrix/strix: AI-Powered Penetration Testing Goes Mainstream
Strix is an open-source AI pen testing tool designed to find and help fix vulnerabilities in applications, trending today in both the overall and Python charts at over 37,000 total stars. It's part of a growing category of AI-augmented offensive security tooling. That projects like this are now standard open-source fare reflects how much the attack surface has expanded in an AI-heavy stack — and how quickly the defensive community is moving to keep up.