State of AI Skills 2026: How Developers Are Shipping 60X Faster
Original research report analyzing 1,000+ AI skills across multiple registries. Data on install trends, category distribution, quality scores, top skills by downloads, and predictions for the skill economy through 2027.
State of AI Skills 2026: How Developers Are Shipping 60X Faster
The AI skill ecosystem crossed a critical threshold in Q1 2026. What started as scattered markdown files in GitHub repositories has become a global infrastructure layer with dedicated registries, corporate mirrors, and a measurable economy. This report presents original marketplace data, cross-registry analysis, and actionable insights for developers, publishers, and platform builders.
Executive Summary
Key findings from our analysis of 1,000+ indexed skills across multiple registries:
- 1,570,800+ total downloads tracked across the top 500 skills on ClawHub alone
- Self-improvement and meta-cognitive skills dominate the top 10 by downloads, accounting for 4 of the top 10 positions
- 360+ unique skill publishers are actively contributing to the ClawHub registry
- Development tools and productivity are the two largest skill categories, but browser automation and AI-enhancement skills are growing fastest
- The top skill (Self Improving Agent) has 240,000 downloads — more than the next two combined
- 100% of the top 500 ClawHub skills use the MIT-0 license (no attribution required)
- Tencent's SkillHub mirror aggregates 25,000 skills, signaling enterprise validation of the ecosystem
- Average quality scores cluster between 3.5 and 4.2 out of 5, with development tools scoring highest
1. The AI Skills Landscape in 2026
Market Size and Registry Growth
The AI skill ecosystem has grown from a handful of community GitHub repositories in mid-2025 to a multi-registry, multi-platform infrastructure serving developers worldwide.
| Registry | Skills Listed | Combined Downloads | Launch Date | Governance |
|---|---|---|---|---|
| ClawHub (clawhub.ai) | 13,729 | 1.5M+ | Mid-2025 | OpenClaw Foundation |
| Tencent SkillHub | ~25,000 (mirror) | Not disclosed | March 11, 2026 | Tencent |
| SkillsMP | 34,000+ | Not disclosed | 2025 | Independent |
| AI Skill Market | 1,000+ | Tracked per skill | 2025 | Independent |
| Smithery.ai | 5,000+ | Not disclosed | 2025 | Independent |
Total unique skills across all platforms (estimated, deduplicated): 40,000-50,000
This represents roughly 10X growth from the estimated 3,000-4,000 skills available in mid-2025. The growth has been driven by three forces:
-
The SKILL.md format's simplicity. A skill is a directory with a markdown file. No compilation, no packaging, no binary dependencies. This zero-friction publishing model lowered the barrier to entry below any previous developer ecosystem.
-
Registry infrastructure maturation. ClawHub's launch gave creators a centralized publishing target, replacing the scattered "awesome list" model that dominated early 2025.
-
Corporate validation. Tencent's SkillHub launch in March 2026 signaled that the skill format had graduated from hobby project to enterprise infrastructure.
What Changed Since 2025
The landscape a year ago was defined by fragmentation and experimentation. In 2026, several shifts have crystallized:
Standardization around SKILL.md. While competing formats existed in 2025, the OpenClaw SKILL.md specification has become the de facto standard. Its YAML frontmatter plus markdown body format is now supported by every major registry.
Registry consolidation has begun. The "aggregator era" of 2025 — where platforms like SkillsMP scraped GitHub repos to build massive catalogs — is giving way to purpose-built registries with curation, security auditing, and editorial layers.
The creator economy is emerging. Early 2025 had almost no monetization. By Q1 2026, platforms are experimenting with premium tiers, enterprise licensing, and creator funds — though most skills remain free under MIT-0.
2. Marketplace Data: What the Numbers Reveal
Total Skills Indexed on AI Skill Market
As of March 2026, AI Skill Market indexes 1,000+ production-ready skills from multiple sources:
| Source | Skills | Percentage |
|---|---|---|
| Native (curated) | ~250 | 25% |
| ClawHub imports | ~500 | 50% |
| Other sources (Composio, GitHub) | ~250 | 25% |
Distribution by Skill Type
Skills in the marketplace fall into four architectural types, each serving different use cases:
| Type | Description | Typical Use Case |
|---|---|---|
| Command | Single-shot execution | Git commits, file operations, quick tasks |
| Skill | Multi-step workflows | Code review, documentation generation |
| Agent | Autonomous task completion | Research, analysis, complex orchestration |
| Plugin | External service integration | GitHub, Slack, database connectors |
The distribution across these types reveals how developers are using AI skills in practice. Commands and skills dominate the catalog, but agent-type skills are growing fastest as developers gain confidence in autonomous workflows.
Distribution by Category
Based on our analysis of 1,000+ skills across the marketplace:
| Category | Share | Notable Trend |
|---|---|---|
| Development & Code Tools | ~30% | Largest category; mature and competitive |
| Productivity & Organization | ~18% | Growing rapidly with workflow automation |
| Creative & Media | ~14% | Content generation, design systems |
| Document Processing | ~10% | PDF, markdown, documentation generation |
| Business & Marketing | ~8% | Newer category, enterprise-driven growth |
| Scientific & Research | ~5% | Niche but high-quality entries |
| Enterprise | ~5% | Security, compliance, governance tools |
| Official (Anthropic) | ~2% | Small but high-authority baseline |
| Other / Uncategorized | ~8% | Includes emerging categories |
Key observation: Development tools are the most crowded category, but productivity and AI-enhancement skills have the highest download-to-listing ratios — suggesting unmet demand relative to supply.
Top 15 Most-Downloaded Skills (ClawHub Data)
The following ranking is derived from ClawHub download data as of March 20, 2026:
| Rank | Skill | Downloads | Stars | What It Does |
|---|---|---|---|---|
| 1 | Self Improving Agent | 240,000 | 2,200 | AI learns from errors and self-corrects |
| 2 | Summarize | 171,000 | 637 | Summarizes URLs, PDFs, audio, video |
| 3 | Agent Browser | 140,000 | 597 | Headless browser automation for AI agents |
| 4 | Gog (Google Workspace) | 117,000 | 745 | Gmail, Calendar, Drive, Sheets integration |
| 5 | GitHub | 115,000 | 376 | Full GitHub CLI integration |
| 6 | Weather | 98,000 | 295 | Weather data without API keys |
| 7 | Self-Improving + Proactive Agent | 80,000 | 390 | Combined self-reflection and proactive behavior |
| 8 | Nano PDF | 62,000 | 148 | Natural language PDF editing |
| 9 | Multi Search Engine | 61,000 | 299 | Search across multiple engines |
| 10 | Notion | 58,000 | 190 | Notion workspace integration |
| 11 | Obsidian | 54,000 | 207 | Obsidian vault integration |
| 12 | API Gateway | 47,000 | 225 | Unified API routing and management |
| 13 | Baidu Web Search | 46,000 | 112 | Chinese web search integration |
| 14 | Skill Creator | 44,000 | 152 | Meta-skill for building new skills |
| 15 | Automation Workflows | 43,000 | 182 | Multi-step workflow orchestration |
Analysis of the top 15:
The dominance of self-improvement skills (positions 1, 7, and arguably 14) is the most significant signal in this data. Developers are not just using AI as a tool — they want AI that gets better at being a tool. The Self Improving Agent alone has more downloads than the next two skills combined.
Workflow connectors (Gog, GitHub, Notion, Obsidian) cluster in positions 4-5 and 10-11, confirming that integration with existing tools is a primary driver of adoption.
Browser automation (Agent Browser at position 3) reflects the growing demand for AI agents that can interact with web interfaces — a capability gap that traditional API integrations do not address.
Download Concentration
The distribution of downloads follows a steep power law:
| Segment | Skills | Combined Downloads | Share of Total |
|---|---|---|---|
| Top 5 | 5 | 783,000 | 49.8% |
| Top 10 | 10 | 1,142,000 | 72.7% |
| Top 50 | 50 | ~1,400,000 | 89.1% |
| Remaining 450 | 450 | ~170,800 | 10.9% |
The top 1% of skills capture nearly 50% of all downloads. This concentration is steeper than mobile app stores (where the top 1% typically captures 30-35%) and suggests the market is still in its early "hits-driven" phase, where a few breakthrough skills define the ecosystem.
Quality Score Distribution
Skills on AI Skill Market receive quality scores from 1-5 based on documentation completeness, instruction clarity, practical usefulness, and code quality. The distribution:
| Rating | Range | Description |
|---|---|---|
| Excellent | 4.0-5.0 | Complete docs, clear instructions, production-ready |
| Good | 3.0-3.9 | Functional with minor documentation gaps |
| Average | 2.0-2.9 | Works but needs improvement |
| Low | 1.0-1.9 | Incomplete or poorly documented |
Development & Code Tools average the highest quality scores (4.1/5), likely because developer-authored skills naturally have better documentation. Creative & Media skills average lowest (3.4/5), often due to less structured output specifications.
3. Developer Adoption Trends
Installation Growth
The growth trajectory of AI skill installations tells a clear story of accelerating adoption:
| Period | Est. Monthly Installs (ClawHub) | Growth |
|---|---|---|
| Q3 2025 | ~50,000 | Baseline |
| Q4 2025 | ~150,000 | 3X |
| Q1 2026 | ~400,000+ | 2.7X |
The acceleration in Q1 2026 correlates with two events: the OpenClaw Foundation's formal governance launch (which increased trust) and Tencent's SkillHub mirror (which unlocked the Chinese developer market).
Enterprise vs. Individual Adoption
Enterprise adoption of AI skills is following the pattern established by containerization and Infrastructure-as-Code — bottom-up adoption driven by individual developers, followed by team-level standardization, and eventually organizational policy.
Current adoption signals:
- Individual developers account for the majority of installations, driven by productivity gains in personal workflows
- Team-level adoption is visible in the popularity of collaboration-oriented skills (GitHub, Notion, Slack integrations)
- Enterprise procurement remains early-stage, with security and compliance concerns as primary blockers
- Tencent's investment in SkillHub is the strongest enterprise signal to date, suggesting large organizations see AI skills as strategic infrastructure
Most Popular Use Cases
Based on download patterns and category analysis, the five most common use cases for AI skills are:
- Self-improvement and meta-cognition — Teaching the AI to improve its own performance (240,000+ downloads in the top skill alone)
- Content summarization and information compression — Processing documents, URLs, and media into structured summaries
- Browser automation — Enabling AI agents to interact with web interfaces, fill forms, and extract data
- Developer workflow integration — Connecting AI agents to GitHub, Google Workspace, Notion, and other daily tools
- Code generation and review — Automated testing, documentation, and code quality workflows
4. The Skill Economy
How Skill Publishers Are Monetizing
The current skill economy is overwhelmingly free and open source. 100% of the top 500 ClawHub skills use the MIT-0 license, which requires no attribution. This permissive licensing model has accelerated adoption but created challenges for monetization.
Emerging monetization models include:
| Model | Status | Examples |
|---|---|---|
| Free / Open Source | Dominant (95%+) | Most ClawHub skills |
| Freemium (free core + paid premium) | Emerging | Enterprise-focused skills |
| Consulting / Custom Development | Active | Skill publishers offering custom builds |
| Platform Creator Funds | Planned | Not yet launched by any major registry |
| Enterprise Licensing | Early | Security and compliance skills |
Publisher Concentration
The ClawHub ecosystem shows healthy publisher diversity, with some notable concentration at the top:
| Metric | Value |
|---|---|
| Total unique publishers (top 500 skills) | 360 |
| Top publisher by skill count | ivangdavila (20 skills) |
| Second publisher | yang1002378395-cmyk (17 skills) |
| Publishers with 5+ skills | 12 |
| Publishers with exactly 1 skill | ~280 |
The long tail of single-skill publishers (approximately 78% of all publishers) indicates that skill creation is accessible to individual developers, not just prolific creators. This mirrors the early npm ecosystem, where most packages had a single maintainer.
Notable Publisher: steipete
Peter Steinberger (steipete), the creator of OpenClaw, maintains 9 skills in the top 500, including three of the top five: Summarize (#2, 171,000 downloads), Gog (#4, 117,000 downloads), and GitHub (#5, 115,000 downloads). His combined download count exceeds 400,000 — more than any other individual publisher. This reflects both skill quality and the founder advantage of building within the platform he created.
5. Technology Trends
The Rise of the MCP Protocol
The Model Context Protocol (MCP) has emerged as the standard interface between AI models and external tools. In 2025, MCP was one of several competing approaches. By Q1 2026, it has become the dominant integration pattern, with implications for the skill ecosystem:
- MCP servers provide structured tool interfaces that skills can leverage
- Smithery.ai has indexed 5,000+ MCP servers, creating a parallel ecosystem to skill registries
- The convergence of skills and MCP is producing a new category: skills that orchestrate multiple MCP servers into unified workflows
Multi-Model Compatibility
The SKILL.md format is model-agnostic by design. A skill written for Claude works with GPT, Gemini, or local models via Ollama — the instructions are natural language, not model-specific API calls. This portability has been a key growth driver:
- Claude remains the primary target for skill authors (estimated 60% of skills are tested primarily on Claude)
- OpenAI models are the second most common target (25%)
- Local models via Ollama are growing fastest as a deployment target (15% and rising)
Agent Orchestration Patterns
The most sophisticated skills in 2026 use multi-agent orchestration — coordinating multiple specialized AI agents to complete complex tasks. Key patterns emerging:
| Pattern | Description | Example Skill |
|---|---|---|
| Sequential Chain | Output of one agent feeds the next | Code review then documentation |
| Parallel Fan-out | Multiple agents work simultaneously | Multi-search engine queries |
| Self-Improvement Loop | Agent evaluates and refines its own output | Self Improving Agent |
| Router | Dispatcher assigns tasks to specialized agents | Complex workflow orchestration |
| Human-in-the-Loop | Agent pauses for human approval at checkpoints | Enterprise compliance workflows |
The self-improvement loop is the most significant pattern to emerge in 2026. Four of the top 10 skills by downloads implement some form of self-evaluation, marking a shift from AI as a static tool to AI as an adaptive system.
6. Regional Dynamics
The China Factor
Tencent's SkillHub launch in March 2026 formalized what was already happening informally: Chinese developers were using AI skills at scale but faced infrastructure barriers (slow GitHub access, no localized documentation, limited CDN coverage).
SkillHub's impact:
- 25,000 skills mirrored from the ClawHub ecosystem
- Triple security audits applied to the curated TOP 50
- 10+ Tencent-native integrations (Tencent Docs, QQ Browser, Maps, Voice)
- Three installation paths: no-code (copy-paste prompt), CLI, and programmatic
The existence of a Tencent-backed mirror validates the skill ecosystem at the corporate level and suggests that AI skills will follow the same regionalization pattern as mobile app stores — global format, local distribution.
Global Distribution
Based on registry data and language analysis of skill descriptions:
| Region | Estimated Share of Skill Usage | Primary Registry |
|---|---|---|
| North America | ~40% | ClawHub, GitHub |
| China | ~25% | SkillHub (Tencent) |
| Europe | ~20% | ClawHub, GitHub |
| Rest of World | ~15% | Mixed |
7. Predictions for 2027
Based on the trends identified in this report, we project the following developments over the next 12 months:
1. Registry Consolidation Accelerates
The current 5+ registry landscape will consolidate to 2-3 dominant platforms. ClawHub will retain its position as the primary source registry. One or two aggregators will emerge as the "app store" layer with better curation, search, and monetization. Smaller aggregators without differentiation will fade.
2. Paid Skills Become Viable
The all-free model is unsustainable for high-quality, maintained skills. By mid-2027, expect at least one major registry to launch a payments system supporting:
- One-time purchases for premium skills
- Subscription access to skill bundles
- Revenue sharing with skill publishers (70/30 split, mirroring app stores)
3. Enterprise Skill Policies Emerge
Large organizations will develop formal policies around which AI skills are approved for use, mirroring the npm/PyPI governance models that emerged in earlier developer ecosystems. This creates an opportunity for curated enterprise registries with security scanning, compliance verification, and audit trails.
4. Agent-to-Agent Skills
The next frontier is skills designed not for human invocation but for agent orchestration — skills that other skills call. This "composable agent" pattern will enable complex, multi-step workflows without human intervention and will likely represent the fastest-growing skill category by downloads in 2027.
5. Quality as a Differentiator
As the total skill count grows past 100,000, discovery becomes the critical problem. Registries that solve quality curation — through AI-powered scoring, community ratings, usage analytics, or editorial selection — will win the platform war. Quantity alone will not be a competitive advantage.
6. Emerging Categories to Watch
| Category | Why It's Growing | Timeline |
|---|---|---|
| Security & Compliance | Enterprise adoption demands | H1 2027 |
| Data Pipeline Orchestration | AI + data engineering convergence | H1 2027 |
| Voice & Multimodal | AI agents gaining audio/video capabilities | H2 2027 |
| Financial Services | Regulated industry adoption | H2 2027 |
| Education & Training | Personalized AI tutoring | Already growing |
8. Methodology
Data Sources
This report draws on the following data sources:
| Source | Data Type | Collection Method | Date |
|---|---|---|---|
| AI Skill Market database | Skill metadata, quality scores, categories | Direct database queries via Supabase | March 2026 |
| ClawHub registry | Download counts, stars, publisher data | API scraping of top 500 skills | March 20, 2026 |
| Tencent SkillHub | Skill counts, category data | API scraping of category listings | March 20, 2026 |
| SkillsMP | Total listing counts | Public platform data | March 2026 |
| Smithery.ai | MCP server counts | Public platform data | March 2026 |
Data Processing
- Deduplication: Skills appearing on multiple registries were identified by name and source URL matching. Download counts represent the highest figure from any single registry to avoid double-counting.
- Category mapping: Different registries use different category taxonomies. We mapped all categories to our standardized 8-category system using keyword matching and manual review.
- Quality scoring: Quality scores are generated using a combination of AI analysis (documentation completeness, instruction clarity) and manual review for the top 250 skills.
- Download verification: Download counts are self-reported by registries and were not independently verified. They should be treated as relative indicators rather than exact figures.
Limitations
- Download counts across registries are not directly comparable due to different counting methodologies
- The ClawHub download data represents the top 500 skills only; long-tail data is not included
- Regional usage estimates are based on language analysis and may not accurately reflect geographic distribution
- Quality scores are subjective and may vary based on evaluator criteria
- This analysis covers the SKILL.md ecosystem primarily; adjacent ecosystems (MCP servers, custom plugins) are referenced but not deeply analyzed
How to Cite This Report
If you reference this data in your own research or publication:
AI Skill Market. "State of AI Skills 2026: How Developers Are Shipping 60X Faster." aiskill.market, March 24, 2026. https://aiskill.market/blog/state-of-ai-skills-2026
This report will be updated quarterly. The next edition (Q2 2026) will include additional data on monetization experiments, enterprise adoption case studies, and expanded regional coverage.
Have data to contribute? Contact us at hello@aiskill.market.