The skills.sh ecosystem is growing fast — hundreds of community-built skills for Claude Code, Cursor, and other AI coding agents. But there's no easy way to tell which skills are worth installing. This dashboard turns raw ecosystem data into actionable signals across six dashboard pages.
Every day, a GitHub Actions workflow scrapes the skills.sh trending page, captures install counts for every listed skill, and enriches each one with GitHub metadata — stars, forks, issues, language, license, topics, and push dates. That daily snapshot is the single data source behind every chart, table, and score on this site.
The data is publicly available in the snapshots/ directory as timestamped JSON files. Nothing is hidden or curated — what the scraper finds is what you see.
"As a solo developer building side projects with Claude Code, I want to quickly find high-quality, well-maintained skills so I can avoid wasting time on abandoned or low-quality ones."
The Scout page is a sortable, filterable leaderboard of every skill in the ecosystem. Each skill gets a Health Score (0–100) and a Freshness badge so you can spot well-maintained skills at a glance. Filter by language, minimum installs, or freshness level. Search by name, owner, or description.
"As a skill author, I want to understand what traits drive adoption so I can optimize my skill's metadata, documentation, and repo health to grow installs."
The Publisher page shows the shape of the ecosystem through four charts: install distribution (the power-law curve), language breakdown, stars vs. installs (they don't correlate the way you'd expect), and repo age vs. installs. A trait correlation table uses Spearman rank correlation to answer "what actually predicts high installs?" A publisher leaderboard shows who's leading in total installs and skill count.
"As a team lead evaluating skills for my engineering team, I want to compare skills side-by-side on maintenance, popularity, and risk signals so I can make informed adoption decisions."
The Compare page lets you search and select 2–4 skills, then see them side-by-side in a radar chart and detailed comparison table. Risk flags surface concerns like missing licenses, stale repos, and high issue ratios so you can make informed decisions before adding a skill to your team's workflow.
"I want to see which skills are gaining momentum and which are fading, so I can spot emerging tools early."
The Trends page tracks ecosystem growth over time using daily snapshots. An install growth chart shows total ecosystem installs, a top movers table highlights skills with the biggest install changes (toggle between total period and last day), and a skill trend lookup lets you search any skill to see its install history with sparklines.
"I want to understand what kinds of skills exist — how many are AI tools vs. frontend helpers vs. DevOps automation — so I can find skills by intent, not just popularity."
The Categories page classifies all 600+ skills into 12 semantic categories (AI & ML, Developer Workflow, Frontend & UI, Cloud & Infrastructure, and more) using Gemini Flash. A horizontal bar chart shows installs per category, and an expandable table lists the top skills within each category. Classifications are cached and re-run automatically when new skills appear.
The Health Score is a weighted composite of three signal groups:
The score is designed to be directionally useful, not definitive. A score of 80 doesn't mean a skill is "good" — it means it has strong signals across the dimensions we can measure from public metadata alone. Always read the actual skill code before installing.
This dashboard is part of a broader longitudinal study of the skills.sh ecosystem, tracking four research questions:
Daily snapshots enable time-series analysis that a single scrape cannot. The Trends page already tracks install growth and top movers, and the Categories page provides semantic classification. Next up: diffusion modeling to identify which traits predict faster adoption.
Snapshots are collected daily. Each snapshot captures install counts and GitHub metadata for all trending skills. The dashboard always loads the most recent snapshot. The date shown in the top-right corner of each page tells you exactly which snapshot you're viewing.
Everything is open source: github.com/vishalsachdev/skills-sh. The scraper is a single Python script, the dashboard is vanilla HTML/CSS/JS with Chart.js, and there is no build step. PRs welcome.