NeverSight/learn-skills.dev

Curated high-quality AI Agent Skills. Search, install, copy and share.
Works with Claude Code, Cursor, OpenClaw, and other AI coding tools.
Web App: https://www.learn-skills.dev — Search, install, copy and share AI Agent Skills.
English | 简体中文 | 繁體中文 |
日本語 |
한국어 |
Français |
Deutsch |
Español |
Italiano |
Русский |
العربية
## Data Sources
### Current Providers
skills.sh - Community-curated skills leaderboard
All Time (/) - Total installs ranking
Trending (/trending) - Recent growth ranking
Hot (/hot) - Daily installs ranking
### Planned Providers
GitHub Trending - Popular skill repos on GitHub
Awesome Lists - Curated awesome- lists for AI agent skills
### Manual Skills
Skills not tracked by any provider can be manually added via data/manual_skills.json:
json
"skills": [
"source": "owner/repo",
"skillId": "skill-name",
"name": "Skill Display Name",
"installs": 1
Manual skills will be:
Fetched for their SKILL.md from GitHub (using standard skill folder detection)
Included in skills_index.json with providerId: "manual"
Not overwritten by the crawler (they persist across runs)
Deduplicated: If skills.sh later tracks a manual skill, it will use skills.sh data instead
Note: installs should be at least 1 (minimum value).
## Output Files
The crawler generates files in the data/ directory:
### data/skills.json
Complete skill data containing all three leaderboards:
json
"updatedAt": "2024-01-27T00:00:00.000Z",
"allTime": [...],
"trending": [...],
"hot": [...]
### data/skills_index.json
Website-friendly index for all skills (built from data/skills.json):
Includes description as a path to description_en.txt (when a cached SKILL.md exists under data/skills-md/)
Includes skillMdPath so your website can fetch and render the full markdown
Deduplicated by id ( / ). If the upstream data contains duplicates, the index keeps the entry with the highest installsAllTime.
### data/feed.json
Simplified feed format (top 50 from each leaderboard).
It also tries to enrich each item with a description by fetching the corresponding GitHub SKILL.md (cached under data/skills-md/):
json
"title": "Skills Feed",
"description": "Aggregated AI agent skills from multiple sources",
"link": "https://github.com/user/skills_feed",
"updatedAt": "2024-01-27T00:00:00.000Z",
"topAllTime": [...],
"topTrending": [...],
"topHot": [...]
### data/skills-md/
Cached SKILL.md files fetched from GitHub, using common skill folder locations such as:
skills/ /SKILL.md (most common)
.claude/skills/ /SKILL.md
.cursor/skills/ /SKILL.md
.codex/skills/ /SKILL.md
plugins/ /skills/ /SKILL.md (common in plugin-based repos, e.g. Expo)
When a SKILL.md is present, the crawler also generates:
description_en.txt (extracted from the SKILL.md frontmatter description when available)
By default, the crawler only fetches SKILL.md for skills included in the top lists (to keep the daily job fast).
If you really want to sync all skills from data/skills.json, you can run:
bash
SYNC_ALL_SKILL_MDS=1 bun run crawl
### data/feed.xml
RSS 2.0 feed (XML). This is meant for RSS readers / subscriptions.