Instead of abstract theory, this course starts with a real analysis of 39.366 skills collected from the skills.sh. You’ll understand the ecosystem, recognize what a skill is good, dissect the best ones and learn to create your own — focusing on what really matters.
A snapshot of the ecosystem that serves as the basis for the course (collected via the skills.sh API):
📊 The data that changes everything: 60% of skills have fewer than 100 installs, and only 131 exceed 100 thousand. The 100 most installed concentrate 43,7% of all installs. The ecosystem follows a power law — and this course teaches you how to play on the right side of the curve.
A journey from overview to practice: understand → recognize → dissect → create → think. Each track ends with the practical trio: ⭐ The Best · 🛠️ How to Create · 🚀 Advanced Tips — 26 modules total, with the final lesson on Regras 2026.
What an agent skill is, why the format won, and the real skills.sh landscape: 39k skills, the power law, and the 16 groups. + Starter Pack · First skill in 10 min · read the ecosystem like a pro.
How to read quality signals (installs, source, description) and a showcase of the best skills by group, with real examples. + the best by signal · create a high-quality skill · signals only experts notice.
Inside SKILL.md: frontmatter, body, progressive disclosure in 3 levels, and the description that makes the skill trigger at the right time. + anatomies to imitate · build a SKILL.md from scratch · multi-file and routing.
From intent to first draft, and the test → evaluate → iterate loop with evals and description optimization. The skill-creator method. + the best meta-skills · full walkthrough with evals · optimization and benchmarking.
The mental models: when a skill is (and ISN’T) worth it, the anti-patterns that kill skills, and how to publish, version, and measure—from GitHub to the portal. + the best process skills · from decision to publish · governance, security, and scale · 2026 rules for 5.5 models.
Trail 2 breaks down why these (and the best in each group) made it there.
More than 100 free courses on AI, agents, and automation. Start here and explore the rest.
View all courses on the portal →