Quality-scored. Not popularity-ranked.

Find AI skills that
actually ship.

The curated directory for AI coding skills โ€” scored on 12 structural quality signals across Claude Code, Cursor, Codex, Windsurf, Cline & more.

0

Skills indexed

0

GitHub stars tracked

0

Skill types

Quality first

Top Rated

View all โ†’

Fresh off the pipeline

Recently Added

View all โ†’

Browse by workflow

Skill Types

Answer guides

The best skill for the job

All guides โ†’

Direct, quality-ranked answers to "what's the best skill for X?" โ€” by agent platform and by job.

Trust & Quality Guide

Every skill is quality-scored based on structural signals โ€” not just GitHub stars.

Quality Tiers

Curated

High-viability skill with gotchas, examples, verification layers.

Promising

Good structure, some quality signals present.

Experimental

Basic skill, limited quality signals.

Risk & Trust

Official

Published by a platform vendor (Anthropic, Cursor, OpenAI, etc.).

Low Risk

Standard permissions. Safe for general use.

Medium Risk

Extended permissions. Review before use.

Maintenance Status

Active

Updated recently

< 90 days

Stable

Feature-complete with releases

> 90 days + releases

Stale

Infrequent updates, no releases

90-365 days

Unmaintained

No recent updates

> 365 days

Unknown

No commit data

N/A

The pipeline

How It Works

01

Multi-Platform Discovery

Skills are discovered across Claude Code, Cursor, Codex, Windsurf, and Cline by scanning for platform-specific markers like SKILL.md, .cursorrules, AGENTS.md, and more.

02

Quality Scoring

Each skill is scored on 12 structural quality signals: gotchas/edge cases, progressive disclosure, verification layers, code examples, and more. Quality is 60% of the ranking โ€” not stars.

03

AI Categorization

Skills are classified into 9 use-case types derived from production lessons scaling to 500+ skills at Uber. Each type maps to a real workflow.

For your agent

Connect over MCP

Query the directory from Claude Code, Cursor, or any MCP client โ€” quality-first search, full skill profiles with risk assessments, collections, and a live demand pulse. Rankings are evidence, not endorsements, and no install commands are served.

$ claude mcp add --transport http skill-of-skills https://skills.911fund.io/api/mcp

Or point any agent at /api/mcp. Full tool + endpoint reference in llms.txt.

search_skills

Quality-first ranked search

get_skill

Full profile + risk reasons

compare_tools

Side-by-side, 2โ€“5 tools

get_collection

Ranked tools by workflow

list_collections

Buyer-intent topics

demand_pulse

K-anonymous demand signal