Top 10 that week
-
6 use cases
REPO
debpalash/VoiceStudioVoice Studio
A desktop app that runs locally for cloning and designing voices, dubbing videos, dictating, transcribing, and making audiobooks, with several speech engines and a local API for agents.
Why it's hot +12,910 stars this week, 96k total · Python
Who should care Creators and developers who want ElevenLabs-style voice tools running on their own hardware.
-
4 use cases
REPO
cloudflare/security-audit-skillSecurity Audit Skill
A skill for coding agents that runs a six-phase security audit of a codebase, using isolated agents to hunt for flaws, try to disprove each candidate, and write verified findings to JSON and markdown reports.
Why it's hot +12,838 stars this week, 55k total · JavaScript
Who should care Developers and security engineers who want a coding agent to audit their own repo with checkable, structured results.
-
3 use cases
REPO
pbakaus/impeccableImpeccable
A design skill pack for coding agents: commands like polish, bolder, critique and audit steer the look of a page, review it for usability and accessibility, and catch generic AI design before it ships.
Why it's hot +11,323 stars this week, 97k total · Rust
Who should care Developers who build front ends with Claude Code, Cursor, Codex or other agents and want less generic looking pages.
-
3 use cases
REPO
HKUDS/DeepTutorDeep Tutor
An open-source AI tutoring app with learning paths, assessments, notebooks, narrated video lessons and tutoring memory, plus a CLI for agents.
Why it's hot +9,440 stars this week, 52k total · Python
Who should care Students, teachers and builders who want a personalized AI tutor they can run themselves.
-
4 use cases
REPO
affaan-m/ECCECC
An add-on pack that gives coding agents memory, security checks and a research-first way of working.
Why it's hot +6,506 stars this week, 294k total · JavaScript
Who should care Developers who want to squeeze more reliable work out of their coding agent.
-
3 use cases
REPO
JetBrains/thinkrailThinkrail
An agentic IDE from JetBrains built around the pi coding agent, with isolated git worktree workspaces, living specs that keep project knowledge, and reusable skills.
Why it's hot +5,495 stars this week, 6.7k total, talked about in 2 AI sources we track · TypeScript
Who should care Developers who want to run several coding agents in parallel on one repo and merge the best branch.
-
3 use cases
REPO
TauricResearch/TradingAgentsTrading Agents
A multi-agent LLM framework where analyst, researcher and manager agents study a stock's news, filings and prices and produce a trading decision, with backtests scored on later closes.
Why it's hot +5,490 stars this week, 119k total, picked up by one AI source we track · Python
Who should care Developers and researchers experimenting with LLM agents for market analysis. Not financial advice: paper trade first.
-
4 use cases
REPO
obra/superpowersSuperpowers
A full working method for coding agents, packaged as skills that make them plan, test and review like a disciplined team.
Why it's hot +5,474 stars this week, 310k total · Shell
Who should care Developers who want their agent to follow a process instead of improvising.
-
3 use cases
REPO
makecindy/cindyCindy
An open-source AI agent app that links your computers and phone, runs Claude Code, Codex or Pi as its engine, and works with your files, browser and apps.
Why it's hot +5,366 stars this week, 8.8k total · TypeScript
Who should care People who use more than one coding agent across several devices and want one interface with memory, skills, MCP and automations.
-
3 use cases
REPO
Leonxlnx/taste-skillTaste Skill
An Agent Skill for Claude Code and Codex that gives coding agents an art direction, real credited pictures and a check in a real browser, so they design pages instead of a generic landing page.
Why it's hot +4,712 stars this week, 107k total · JavaScript
Who should care Developers who build websites or app screens with coding agents. It needs Node 22+ and Chrome, Edge or Chromium.
Past weeks
How we rank
Every day we snapshot star counts for a few hundred repos from GitHub's trending lists and the newest AI projects, keep only AI projects, and score each one on star growth, how many of the AI newsletters, forums and feeds we read are talking about it, and how new it is. Then we read each README and sum it up in one plain sentence. Star farms and sketchy projects get dropped.