--- name: agent-skills-search description: Search and summarize skills from multiple directories (agent-skills.cc + skills.sh) using public sitemaps + page fetch (no APIs). Use when the user asks to find relevant agent skills, wants top N matches for keywords, or wants links + quick summaries (e.g., “top 8 skills for Laravel\/Rust\/DevOps\/testing\/security”). --- # Agent Skills Search Find relevant skills across: - **https:\/\/agent-skills.cc** - **https:\/\/skills.sh** …and return a clean **Top 8** list with links and one-line summaries. This skill avoids fragile client-side search scraping and avoids private APIs. Instead it uses: - `https:\/\/agent-skills.cc\/sitemap.xml` for discovery - `https:\/\/skills.sh\/sitemap.xml` for discovery - fetching individual skill pages for details ## Quick workflow (Top 8, sitemap-first, multi-source) 1) **Get the query** - Ask for 1–5 keywords. - Default output size: **Top 8**. 2) **Discover candidate skill pages via sitemaps** - Fetch both sitemaps: - `https:\/\/agent-skills.cc\/sitemap.xml` - `https:\/\/skills.sh\/sitemap.xml` - Extract candidate URLs: - agent-skills.cc: `https:\/\/agent-skills.cc\/skills\/` - skills.sh: `https:\/\/skills.sh\/\/\/` - Filter candidates by keyword match (case-insensitive) in the URL path first (fast). - If too few results, fetch some candidate pages and match against title\/description too. 3) **Fetch details for the best candidates** For each candidate skill page, extract (whatever is present): - Title - Short description - Source\/outbound link(s) (GitHub\/docs) - Popularity signals if visible (stars\/likes\/installs) - Install hint if obvious - skills.sh often uses: `npx skills add ` 4) **Rank + dedupe (cross-source)** - Remove duplicates: - Prefer dedupe by canonical GitHub `owner\/repo` when you can extract it. - Fallback: title+description similarity. - Prefer strong relevance: 1) keyword match in title \/ slug 2) match in description 3) match in tags\/category 4) popularity as tiebreaker 5) **Return Top 8 in a consistent format** - **1. ** — - Page: https:\/\/agent-skills.cc\/skills\/ - Source: - Link(s): ## Useful browsing URLs Use these when the user wants discovery without a keyword query: **agent-skills.cc** - Home: https:\/\/agent-skills.cc\/ - Claude skills: https:\/\/agent-skills.cc\/claude-skills - Codex skills: https:\/\/agent-skills.cc\/codex-skills - Cursor skills: https:\/\/agent-skills.cc\/cursor-skills **skills.sh** - Home: https:\/\/skills.sh\/ - Sitemap: https:\/\/skills.sh\/sitemap.xml ## Bundled resources ### scripts\/ - `scripts\/search_skills.py` - End-to-end local search helper (both sources): sitemap discovery → fetch details → rank+dedupe → print Top N. - Example: - `python3 scripts\/search_skills.py rag vector --limit 8 --fetch 40` - `scripts\/extract_skill_urls_from_sitemap.py` - Legacy helper (agent-skills.cc only). - `scripts\/extract_skill_urls_from_sitemaps.py` - Extract skill URLs from: - agent-skills.cc sitemap - skills.sh sitemap - Useful for quick local filtering and debugging. ### references\/ If you need URL patterns and extraction guidance, read: - `references\/agent-skills-site-guide.md` - `references\/skills-sh-site-guide.md`