#1: ayghri/i-have-adhd
Python · 44,740 stars total, +16,740 this period
i-have-adhd is a skill for AI coding agents with one job: stop the agent from burying the answer. Coding agents tend to wrap the thing you asked for in long explanations; this skill makes them lead with the answer and keep the output short and scannable, which the author describes as ADHD-friendly output. It gained 16,740 stars this period for a total of 44,740, so the complaint is clearly a common one.
#2: affaan-m/ECC
JavaScript · 257,944 stars total, +7,264 this period
The affaan-m/ECC repository, written in JavaScript, presents itself as an agent harness performance optimization system. It aims to enhance the efficiency and effectiveness of various systems, including Claude Code, Codex, Opencode, and Cursor, among others. The project focuses on leveraging skills, instincts, memory, and research-first development to address performance bottlenecks and improve overall system responsiveness and reliability. Given its significant star count of 257,944 and the addition of 7,264 stars in the recent period, it appears to be gaining traction rapidly. This rapid growth suggests that the project has addressed a pressing need or solved a critical problem within the community, making it a notable addition to the "Cool GitHub Reviews" section.
#3: openai/plugins
JavaScript · 6,654 stars total, +1,181 this period
openai/plugins is OpenAI's curated collection of example plugins for Codex, its coding agent. Each plugin lives in its own folder with a manifest and optional skills, MCP configuration, commands and hooks. The highlighted examples include plugins for Figma, Notion, and building iOS, macOS and web apps. It is a reference for people writing their own Codex plugins as much as a set to install. It has 6,654 stars, 1,181 of them gained this period.
#4: bilawalsidhu/gods-eye-view
JavaScript · 32,563 stars total, +12,931 this period
gods-eye-view is described as a spy-satellite simulator in your browser, except the data is real. It shows live public data on a photorealistic 3D globe: aircraft, ships, satellites, earthquakes, traffic and public cameras, with voice control through a realtime AI agent. You run it locally (or install it through Pinokio); it starts without API keys, and optional keys add more sources.
It gained 12,931 stars this period for a total of 32,563, helped by the creator's video series of the same name.
#5: mksglu/context-mode
TypeScript · 22,717 stars total, +2,102 this period
context-mode is a context-window optimiser for AI coding agents. Its main trick is sandboxing tool output so that only the useful part reaches the model; the project claims a 98% reduction in tool output. It also persists session memory so an agent can pick up where it left off, and it enforces its routing across many agent platforms using MCP (the Model Context Protocol) plus hooks.
It has 22,717 stars, with 2,102 gained this period.