Week of May 9 – May 15, 2026
Published: May 17, 2026
Data Source: star-history.com
Coverage: Weekly trending repositories — 7-day star gain rankings
Author: CodeWanderer
Tags: GitHub | Open Source | AI Agents | Developer Tools | Trending
Table of Contents
- Executive Summary
- The Big Picture: A Week Dominated by AI Agent Skills
- Top 20 Weekly Trending Repositories
- Deep Dives: Top Repository Highlights
- Thematic Analysis: What's Driving the Trend
- Organization Breakdown
- Key Statistics
- Notable Entries & Rising Stars
- What This Means for the Ecosystem
- Conclusion
1. Executive Summary
This week's GitHub trending data (May 9 – May 15, 2026) paints a remarkably clear picture: AI agent skills frameworks have taken over the open-source ecosystem. Of the top 20 fastest-growing repositories, at least 5 contain the word "skills" in their names, and the broader Claude Code / AI coding assistant ecosystem accounts for 3+ additional entries. The week's top repository — mattpocock/skills — accumulated +1,618 stars in a single week, and the combined top 20 repositories collectively gained nearly 14,600 stars in 7 days.
The entry threshold to the top 20 this week is +415 stars, an exceptionally high bar that signals extraordinary developer activity. The trend is unmistakable: the community is racing to build, share, and refine the "skills" (i.e., behavioral patterns, workflows, and tooling) that make AI coding agents genuinely useful in real engineering environments.
2. The Big Picture
┌─────────────────────────────────────────────────────────────────────────────┐
│ GitHub Weekly Trending — May 9–15, 2026: Theme Distribution │
├─────────────────────────────────────────────────────────────────────────────┤
│ │
│ AI Agent Skills Frameworks ████████████████████░░░░░░░ 5 repos (25%) │
│ │
│ Claude Code Ecosystem ██████████████░░░░░░░░░░░░░░░ 3 repos (15%) │
│ │
│ General AI / LLM Tools ████████████░░░░░░░░░░░░░░░░ 3 repos (15%) │
│ │
│ Finance / Trading AI ████████░░░░░░░░░░░░░░░░░░░░ 2 repos (10%) │
│ │
│ Design / Creative AI ████████░░░░░░░░░░░░░░░░░░░░░ 2 repos (10%) │
│ │
│ Developer Tooling / Other ██████████████████████████░░░ 5 repos (25%) │
│ │
└─────────────────────────────────────────────────────────────────────────────┘
This week's trending list is overwhelmingly concentrated in AI agent tooling, with a specific focus on skills-based frameworks — repositories that package reusable workflows, coding patterns, and operational guardrails for AI coding agents like Claude Code, Codex, and similar tools.
The "Skills" Pattern Goes Viral
The most remarkable phenomenon this week is the proliferation of the "skills" pattern for AI agents. Inspired by Matt Pocock's original mattpocock/skills repository (which open-sourced his personal .claude directory), developers worldwide are now creating their own skills collections, specialized skill frameworks, and derivative works:
| Repository | Theme |
|---|---|
mattpocock/skills |
Real engineering skills for Claude Code — TDD, guardrails, debugging |
multica-ai/andrej-karpathy-skills |
Skills based on Andrej Karpathy's AI engineering philosophy |
addyosmani/agent-skills |
Agent skills by Google's Addy Osmani |
anthropics/skills |
Anthropic's official skills framework |
anthropics/financial-services |
Domain-specific AI agent toolkit (finance) |
This is not a coincidence — it reflects a paradigm shift in how developers are thinking about AI coding assistants. Rather than treating AI agents as black-box code generators, the community is now actively engineering the context, constraints, and workflows that make AI agents reliable in production environments.
3. Top 20 Weekly Trending Repositories
3.1 Full Rankings Table
| Rank | Change | Repository | Stars Gained | Owner Type |
|---|---|---|---|---|
| 1 | — | mattpocock/skills | +1,618 | Individual |
| 2 | ▲ +3 | NousResearch/hermes-agent | +1,332 | Research Lab |
| 3 | ▲ +7 | multica-ai/andrej-karpathy-skills | +1,117 | Startup |
| 4 | ▼ -2 | anthropics/financial-services | +1,075 | Company |
| 5 | ▲ +12 | obra/superpowers | +951 | Individual |
| 6 | ▼ -1 | Hmbown/DeepSeek-TUI | +881 | Individual |
| 7 | 🆕 New | farion1231/cc-switch | +866 | Individual |
| 8 | ▲ +10 | affaan-m/everything-claude-code | +858 | Individual |
| 9 | ▼ -4 | addyosmani/agent-skills | +840 | Individual (Google) |
| 10 | — | github/spec-kit | +736 | Organization |
| 11 | ▲ +2 | datawhalechina/hello-agents | +652 | Organization |
| 12 | ▲ +5 | VoltAgent/awesome-design-md | +637 | Organization |
| 13 | ▲ +18 | tinyhumansai/openhuman | +629 | Startup |
| 14 | ▼ -5 | garrytan/gstack | +588 | Individual (YC) |
| 15 | ▲ +8 | ruvnet/ruflo | +502 | Individual |
| 16 | ▼ -7 | TauricResearch/TradingAgents | +497 | Research Lab |
| 17 | ▲ +3 | anthropics/skills | +439 | Company |
| 18 | 🆕 New | anomalyco/opencode | +430 | Company |
| 19 | ▼ -6 | bytedance/UI-TARS-desktop | +428 | Corporation |
| 20 | 🆕 New | yikart/AiToEarn | +415 | Individual |
Legend: ▲ = moved up, ▼ = moved down, — = no change, 🆕 = new entry
3.2 Rank Movement Summary
| Movement Type | Count |
|---|---|
| Rising (▲) | 8 repos |
| Falling (▼) | 6 repos |
| Steady (—) | 5 repos |
| New Entry (🆕) | 3 repos |
3.3 Distribution by Owner Type
| Owner Type | Count | Notable Examples |
|---|---|---|
| Individual developers | 11 | mattpocock, Hmbown, farion1231, affaan-m |
| Companies | 3 | Anthropic, ByteDance, Anomaly Innovations |
| Research labs | 3 | NousResearch, TauricResearch, multica-ai |
| Organizations | 3 | GitHub, DatawhaleChina, VoltAgent |
4. Deep Dives: Top Repository Highlights
4.1 #1 — mattpock/skills (+1,618 stars)
"Skills for Real Engineers. Straight from my .claude directory."
mattpocock/skills is the crown jewel of this week's trending list. Created by Matt Pocock — a renowned TypeScript expert and former Vercel developer — this repository open-sources his personal .claude directory, which contains the agent skills he uses daily for real engineering work.
Why it matters: The repository represents a new paradigm in AI-native knowledge management. Instead of treating AI coding assistants as one-size-fits-all code generators, Pocock has systematically engineered:
- TDD (Test-Driven Development) skills: Instructing AI agents to write tests before implementation
-
Guardrail skills: Blocking dangerous commands (e.g.,
git push --force,rm -rf) with confirmation hooks - Structured debugging skills: Guiding agents through systematic debugging workflows
- TypeScript-specific patterns: Deep integration with TypeScript's type system for better AI output
The repository has spawned numerous derivatives and inspired an entire ecosystem of "skills" repositories this week, making it arguably the most influential open-source project of the week.
4.2 #2 — NousResearch/hermes-agent (+1,332 stars)
"The agent that grows with you."
Hermes Agent is an open-source autonomous AI agent framework developed by Nous Research, released in February 2026. With over 105,000 total stars as of this week, it represents the flagship open-source AI agent project.
Key features:
- Self-improving memory: The agent learns from interactions and becomes more capable over time
- Cross-platform: Runs on Linux, macOS, and with limitations on Windows
- Python-based: MIT-licensed, highly extensible architecture
- v0.11.0 "The Interface Release": Recent major update with 1,556 commits merged
The project represents the "agent that grows with you" philosophy — rather than stateless, single-session AI tools, Hermes Agent maintains persistent context and memory that compounds over time, making it increasingly effective for long-term engineering tasks.
4.3 #3 — multica-ai/andrej-karpathy-skills (+1,117 stars)
"Andrej Karpathy's AI engineering skills, packaged for agents."
This repository applies the skills pattern pioneered by Matt Pocock specifically to the philosophy and techniques of Andrej Karpathy, one of the most influential figures in AI/ML education.
The repository packages Karpathy's documented AI engineering workflows, including:
- Neural network implementation patterns
- Training and fine-tuning workflows
- LLM utilization best practices
- Reproducibility and experiment tracking
Why it matters: This represents the generalization of the skills pattern — it's not just about coding tools, but about encoding the workflows and mental models of expert practitioners into reusable agent behaviors.
4.4 #4 — anthropics/financial-services (+1,075 stars)
Anthropic's official domain-specific agent toolkit for financial services.
This is Anthropic's second entry in this week's top 20 (alongside anthropics/skills at #17). The financial-services repository provides a specialized toolkit for deploying Claude-powered agents in financial services contexts, including:
- Compliance-aware workflows
- Risk assessment patterns
- Regulatory document processing
- Audit trail generation
Significance: Anthropic's dual presence in the top 20 demonstrates the company's dual-track strategy — both a general-purpose skills framework (anthropics/skills) and domain-specific solutions (anthropics/financial-services).
4.5 #7 — farion1231/cc-switch (+866 stars, New Entry)
Claude Code environment switching tool.
A fresh new entry this week, cc-switch addresses a practical pain point: developers who work with multiple Claude Code environments (e.g., different projects, different API keys, different configurations) need a way to quickly switch between them without reconfiguring from scratch.
This tool reflects the maturation of the Claude Code ecosystem — as adoption grows, the tooling around it (configuration managers, switchers, context managers) is proliferating rapidly.
4.6 #8 — affaan-m/everything-claude-code (+858 stars)
"A comprehensive resource collection for Claude Code users."
Another Claude Code ecosystem entry, everything-claude-code serves as a curated knowledge base aggregating tips, tricks, workflows, integrations, and community discoveries around Anthropic's Claude Code product. Think of it as a comprehensive field guide for Claude Code power users.
4.7 #10 — github/spec-kit (+736 stars)
"Toolkit to help you get started with Spec-Driven Development."
GitHub's official spec-kit is an interesting outlier in the top 10 — it's not AI-specific per se, but represents a systematic approach to AI-augmented development. Spec-Driven Development (SDD) is a methodology that structures AI-assisted coding through a specification → plan → task → implementation workflow:
┌─────────────────────────────────────────────────────────────┐
│ Spec-Driven Development Workflow │
├─────────────────────────────────────────────────────────────┤
│ │
│ Step 1: SPECIFICATION │
│ └─ Write natural-language requirements ("what" & "why") │
│ │ │
│ ▼ │
│ Step 2: PLAN │
│ └─ AI breaks down spec into implementable tasks │
│ │ │
│ ▼ │
│ Step 3:TASKS │
│ └─ Structured, reviewable task list │
│ │ │
│ ▼ │
│ Step 4: IMPLEMENTATION │
│ └─ AI writes code based on the plan, not the prompt │
│ │
└─────────────────────────────────────────────────────────────┘
Why it's trending: Developers are increasingly frustrated with "vibe coding" (AI-assisted coding without structure). Spec-kit provides a disciplined framework that keeps AI agents grounded in well-defined requirements rather than generating ad-hoc solutions.
5. Thematic Analysis
5.1 The "Skills Pattern" Explained
The most significant trend this week is the emergence and rapid adoption of the "skills pattern" for AI coding agents. Conceptually, a "skill" is a structured configuration file or script that defines how an AI agent should behave in specific situations.
┌─────────────────────────────────────────────────────────────────────────┐
│ AI Agent Skills Architecture │
├─────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌───────────────┐ ┌───────────────┐ ┌───────────────┐ │
│ │ AI Coding │ │ Skills │ │ Output │ │
│ │ Agent │◄────│ (Config & │────►│ (Constrained│ │
│ │ (Claude, │ │ Patterns) │ │ + Reliable)│ │
│ │ Codex) │ │ │ │ │ │
│ └───────────────┘ └───────────────┘ └───────────────┘ │
│ │
│ Skills Examples: │
│ ├─ Guardrails: Block dangerous git commands │
│ ├─ TDD patterns: Require tests before code │
│ ├─ Debug workflows: Structured error investigation │
│ ├─ Domain patterns: TypeScript, React, Python best practices │
│ └─ Review checklists: PR description templates, code style guides │
│ │
└─────────────────────────────────────────────────────────────────────────┘
Why the skills pattern is gaining traction:
| Driver | Explanation |
|---|---|
| Reliability | Raw AI agents make mistakes; skills constrain behavior to proven patterns |
| Reproducibility | Teams can share and version-control their AI workflows |
| Onboarding | New team members inherit best practices automatically |
| Domain expertise | Encoding expert workflows makes junior devs more effective |
| Safety | Guardrails prevent catastrophic actions (data loss, security issues) |
5.2 The Claude Code Ecosystem Explosion
An interesting sub-trend this week is the rapid growth of the Claude Code ecosystem. With 3 repositories directly related to Claude Code (cc-switch, everything-claude-code, spec-kit which integrates with it), and additional skills repositories explicitly designed for Claude Code (mattpocock, anthropics, addyosmani), the Anthropic coding agent is generating significant community tooling activity.
Claude Code ecosystem map:
┌─────────────────────────────────────────────────────────────────────────┐
│ Claude Code Ecosystem (Week of May 9–15, 2026) │
├─────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌─────────────────┐ │
│ │ Claude Code │ │
│ │ (Core Product) │ │
│ └────────┬────────┘ │
│ │ │
│ ┌───────────────────────┼───────────────────────┐ │
│ │ │ │ │
│ ▼ ▼ ▼ │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ cc-switch │ │ everything- │ │ spec-kit │ │
│ │ (Env Manager)│ │ claude-code │ │ (SDD Flow) │ │
│ └─────────────┘ │ (Guide) │ └──────┬──────┘ │
│ └─────────────┘ │ │
│ ┌───────────────────────────────────────────┘ │
│ │ │
│ ▼ │
│ ┌─────────────────────────────────────────────────────────────┐ │
│ │ Skills Repositories (for Claude Code) │ │
│ │ │ │
│ │ mattpocock/skills addyosmani/agent-skills │ │
│ │ anthropics/skills multica-ai/andrej-karpathy-skills │ │
│ └─────────────────────────────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────────────┘
5.3 Domain-Specific AI Agents
Another notable trend: domain-specific AI agent repositories are gaining traction alongside general-purpose tools:
| Repository | Domain | Focus |
|---|---|---|
anthropics/financial-services |
Finance | Compliance, risk, audit trails |
TauricResearch/TradingAgents |
Trading | AI-powered trading strategies |
bytedance/UI-TARS-desktop |
UI Automation | Desktop UI agent by ByteDance |
ruvnet/ruflo |
Flow/Productivity | Workflow automation |
VoltAgent/awesome-design-md |
Design | Design system documentation |
This signals that AI agents are maturing beyond general coding assistance into specialized tools for vertical domains — a natural evolution as the underlying models become more capable and the community develops domain-specific fine-tuning and prompting strategies.
6. Organization Breakdown
6.1 Corporate Presence
| Organization | Repositories | Combined Stars | Strategy |
|---|---|---|---|
| Anthropic |
financial-services, skills
|
~1,514 | Dual-track: general + domain-specific |
| ByteDance | UI-TARS-desktop |
428 | Desktop UI automation |
| GitHub | spec-kit |
736 | Spec-Driven Development tooling |
| Anomaly Innovations |
opencode (new) |
430 | Open-source coding agent |
6.2 Research Lab Presence
| Organization | Repository | Total Stars | Focus |
|---|---|---|---|
| NousResearch | hermes-agent |
~105,000 | Self-improving AI agents |
| TauricResearch | TradingAgents |
(growing) | AI trading systems |
| multica-ai | andrej-karpathy-skills |
(growing) | Expert workflow encoding |
6.3 Community-Driven Repositories
11 out of 20 top repositories this week are owned by individual developers, underscoring the grassroots nature of the AI agent skills movement. This is reminiscent of the early days of npm packages or Vue.js — a single influential individual's idea (mattpocock/skills) has catalyzed a community-wide movement in a matter of weeks.
7. Key Statistics
7.1 Aggregate Numbers
| Metric | Value |
|---|---|
| Total stars (Top 20) | 14,591 stars |
| Average stars per repo | ~730 stars |
| Highest weekly gain |
1,618 stars (mattpocock/skills) |
| Lowest in Top 20 |
415 stars (yikart/AiToEarn) |
| New entries |
3 repos (cc-switch, opencode, AiToEarn) |
| Entry threshold | +415 stars (exceptionally high) |
7.2 Week-over-Week Comparison
| Metric | This Week (May 9–15) | Previous Week | Change |
|---|---|---|---|
| #1 stars gained | 1,618 | ~1,200 (est.) | ▲ +35% |
| #20 threshold | 415 | ~300 (est.) | ▲ +38% |
| AI agent repos in Top 20 | 12+ | ~8 (est.) | ▲ +50% |
| "Skills" repos in Top 20 | 5 | ~2 (est.) | ▲ +150% |
7.3 Historical Context
Comparing this week to recent weeks (based on available data), the "skills pattern" trend is accelerating rapidly:
| Week | # of "Skills" repos in Top 20 | Top repo stars |
|---|---|---|
| 4 weeks ago | ~1 | ~900 |
| 3 weeks ago | ~2 | ~1,000 |
| 2 weeks ago | ~3 | ~1,100 |
| Last week | ~4 | ~1,300 |
| This week | 5 | 1,618 |
8. Notable Entries & Rising Stars
8.1 New Entries (🆕)
Three repositories made their debut in this week's top 20:
| Repository | # | Stars | Why It Matters |
|---|---|---|---|
| farion1231/cc-switch | 7 | +866 | Addresses Claude Code multi-environment management |
| anomalyco/opencode | 18 | +430 | Open-source coding agent from Anomaly Innovations |
| yikart/AiToEarn | 20 | +415 | AI monetization/earnings platform |
8.2 Biggest Jumpers
| Repository | Previous Rank | Current Rank | Change |
|---|---|---|---|
obra/superpowers |
#17 | #5 | ▲ +12 |
tinyhumansai/openhuman |
#31 | #13 | ▲ +18 |
farion1231/cc-switch |
🆕 | #7 | 🆕 +13 (new) |
affaan-m/everything-claude-code |
#18 | #8 | ▲ +10 |
ruvnet/ruflo |
#23 | #15 | ▲ +8 |
obra/superpowers is particularly noteworthy — jumping 12 positions to #5 with +951 stars suggests it offers something the community urgently needs. Based on available data, superpowers appears to be a comprehensive AI/coding superpowers toolkit that aggregates multiple skills and workflows into a cohesive framework.
8.3 Falling Stars
| Repository | Previous Rank | Current Rank | Change | Analysis |
|---|---|---|---|---|
Hmbown/DeepSeek-TUI |
#5 | #6 | ▼ -1 | Minor drop, still strong |
anthropics/financial-services |
#2 | #4 | ▼ -2 | Normal fluctuation |
TauricResearch/TradingAgents |
#9 | #16 | ▼ -7 | Saturated market? |
garrytan/gstack |
#9 | #14 | ▼ -5 | YC backing, stable |
bytedance/UI-TARS-desktop |
#13 | #19 | ▼ -6 | Desktop UI is niche |
9. What This Means for the Ecosystem
9.1 The Skills Pattern is a Paradigm Shift
The most significant takeaway from this week's data is that the "skills pattern" represents a genuine paradigm shift in how developers approach AI coding assistants:
Old paradigm: "Here's a prompt, generate code."
New paradigm: "Here are structured skills (guardrails, workflows, patterns), apply them intelligently."
This shift has profound implications:
- AI agents become trainable: Just as developers train ML models, teams can now train AI agents through skills repositories
- Knowledge becomes portable: Expert workflows encoded in skills can be shared, reviewed, and version-controlled like code
- Safety improves: Guardrail skills prevent catastrophic mistakes, making AI agents safer for production use
- Teams converge on best practices: Shared skills repositories create organizational standards for AI-assisted development
9.2 The Ecosystem is Consolidating Around Claude Code
While the AI coding assistant market includes multiple players (GitHub Copilot, Cursor, Cody, etc.), this week's data suggests the open-source community is consolidating around Claude Code:
- 5 skills repositories are explicitly designed for Claude Code
-
spec-kit(GitHub's official tool) has first-class Claude Code integration - 3 repositories directly address Claude Code ecosystem needs
-
mattpocock/skills— the week's #1 repo — was built for Claude Code
9.3 Domain Specialization is Accelerating
The presence of domain-specific repositories (financial-services, TradingAgents, UI-TARS-desktop) signals that AI agents are moving beyond general coding into vertical domains. This mirrors the evolution of SaaS — from horizontal productivity tools to specialized industry solutions.
9.4 Predictions for Next Week
Based on current momentum, here are educated predictions for next week's trending list:
| Prediction | Confidence | Rationale |
|---|---|---|
| "Skills" repos will exceed 6 in Top 20 | High | Momentum is accelerating |
mattpocock/skills will stay in Top 5 |
High | Community interest still growing |
| New Claude Code ecosystem tools will appear | High | Clear demand signal |
| First "skills marketplace" concept may emerge | Medium | Natural next step |
| Corporate repos (Google, Microsoft) may enter | Medium | They have similar offerings in development |
10. Conclusion
This week's GitHub trending data tells a compelling story: the open-source community has found a new obsession, and it's called "skills."
What started as Matt Pocock sharing his personal .claude directory has exploded into a week-long frenzy of skills repository creation, with at least 5 skills-focused repositories earning spots in this week's top 20 — including one that gained +1,618 stars in 7 days, making it one of the fastest-growing repositories on all of GitHub.
The implications extend far beyond GitHub stars. The "skills pattern" represents a fundamental shift in how we think about AI coding assistants — from stateless code generators to behaviorally-defined agents that can be engineered, shared, reviewed, and improved like any other software artifact.
┌─────────────────────────────────────────────────────────────────────────────┐
│ Key Takeaways │
├─────────────────────────────────────────────────────────────────────────────┤
│ │
│ 🔥 The "Skills Pattern" is viral — 5 skills repos in Top 20 this week │
│ │
│ 🤖 AI agents are maturing — from code generation to structured workflows │
│ │
│ 💼 The ecosystem is consolidating — Claude Code has the most momentum │
│ │
│ 🏢 Domain specialization is accelerating — finance, trading, design agents │
│ │
│ 👥 Community-driven innovation — 11/20 repos are individual projects │
│ │
│ 📈 Entry threshold hit +415 stars — one of the highest ever recorded │
│ │
└─────────────────────────────────────────────────────────────────────────────┘
Whether you're a seasoned developer or just getting started with AI coding tools, this week's trending data offers a clear signal: the future of AI-assisted development is skills-based, structured, and community-driven.
Appendix: Resources & Links
Trending Repositories
Data Sources
- Star History: star-history.com
- GitHub Trending: github.com/trending
- Coding AI Leaderboard: star-history.com/coding-ai-leaderboard
Tools Mentioned
| Tool | Description |
|---|---|
| Claude Code | Anthropic's AI coding assistant |
| Hermes Agent | NousResearch's self-improving AI agent |
| Spec-Kit | GitHub's Spec-Driven Development toolkit |
| DeepSeek-TUI | Terminal UI for DeepSeek models |
| opencode | Anomaly Innovations' open coding agent |
About the Author: 码流怪侠 — CSDN technical blogger, researcher in video codecs, video quality assessment, and AI tools. This analysis is based on publicly available GitHub star history data from star-history.com for the week of May 9–15, 2026.
Disclaimer: Star counts and repository details are sourced from star-history.com and may not reflect real-time data. The analysis and interpretations represent the author's views based on available information.
License: This article is available for redistribution and reference with attribution.
Top comments (0)