Ranked list
Best Claude skills in 2026
“Best” depends on the job, so this page is explicit about the criteria: install count from the public registry, a readable source file, and a security grade. Use it as a shortlist, not a verdict.
How this list is built
- Installs come from the public skills.sh registry, refreshed Nov 17, 2026. Popularity is not quality, but it is the only cross-skill signal that exists.
- Source must be a public repository with a readable
SKILL.md. Every row links to it. - Grade is our static security scan of that file — A means no risky patterns were found in the text.
- No paid placement. Nothing here is sponsored, and we do not claim to have tested these skills.
Top 20 by installs
| # | Skill | Author | Installs | Grade |
|---|---|---|---|---|
| 1 | find-skills | Vercel | 4.4M | A |
| 2 | hyperframes-cli | HeyGen | 1.6M | A |
| 3 | grill-me | Matt Pocock | 1.6M | A |
| 4 | hyperframes | HeyGen | 1.6M | A |
| 5 | agent-browser | Vercel | 1.5M | A |
| 6 | media-use | HeyGen | 1.5M | A |
| 7 | hyperframes-registry | HeyGen | 1.3M | A |
| 8 | grill-with-docs | Matt Pocock | 1.3M | A |
| 9 | improve-codebase-architecture | Matt Pocock | 1.3M | A |
| 10 | tdd | Matt Pocock | 1.2M | A |
| 11 | setup-matt-pocock-skills | Matt Pocock | 1.2M | A |
| 12 | handoff | Matt Pocock | 1.1M | A |
| 13 | frontend-design | Anthropic | 1.1M | A |
| 14 | grilling | Matt Pocock | 1.1M | A |
| 15 | prototype | Matt Pocock | 1.1M | A |
| 16 | triage | Matt Pocock | 1.1M | A |
| 17 | domain-modeling | Matt Pocock | 988K | A |
| 18 | codebase-design | Matt Pocock | 960K | A |
| 19 | teach | Matt Pocock | 957K | A |
| 20 | diagnosing-bugs | Matt Pocock | 940K | A |
Best pick in each category
Coding — find-skills (4.4M, grade A), grill-me (1.6M, grade A) · browse coding
Design — frontend-design (1.1M, grade A), web-design-guidelines (869K, grade A) · browse design
Cloud — microsoft-foundry (726K, grade A), entra-app-registration (713K, grade A) · browse cloud
Video — hyperframes-cli (1.6M, grade A), hyperframes (1.6M, grade A) · browse video
Productivity — lark-doc (497K, grade A), lark-base (494K, grade A) · browse productivity
Marketing — copywriting (257K, grade A), marketing-psychology (179K, grade A) · browse marketing
Database — supabase-postgres-best-practices (488K, grade A), prisma-client-api (432K, grade A) · browse database
SEO — seo-audit (257K, grade A), ai-seo (166K, grade A) · browse seo
Browser — agent-browser (1.5M, grade A), just-scrape (245K, grade A) · browse browser
Document — pptx (257K, grade A), pdf (231K, grade A) · browse document
Why a “best” list is always partial
- Your job is specific. A skill with 4M installs for finding other skills does nothing for a PDF task.
- Install counts are cumulative. A skill can be popular and abandoned; check the repo’s last commit before you rely on it.
- Grades are text scans. They catch risky patterns, not bad instructions. Read the file.
How to evaluate a skill in two minutes
- Open the
SKILL.mdand read thedescription— does it match your task? - Skim the body for steps you would not want an agent to take.
- Check the security grade and what triggered it.
- Install it, run it once on a real task, and compare with the result you get without it.
Frequently asked
What is the single best Claude skill? There is no such thing. The most-installed skill in this index is find-skills, which helps you find other skills — useful, but not a substitute for the one you actually need.
Are these officially recommended by Anthropic? No. This is an independent index; Anthropic publishes its own reference skills, which are included and marked by author.
How often is this updated? Install counts are refreshed with each index build; the date is shown at the top of the page.
Community notes
Found a broken install? Know a skill we missed?
One line is enough: the skill name, what happened, and the command you ran. We read every note and fix the index.