Claude Skills Marketplace

Claude Skills are reusable AI capabilities - single .md or .yaml files that live in your project and tell Claude exactly how to do a specific job. Install once, use everywhere. No prompt engineering on repeat.

Skills work with Claude Code, Cursor, Codex, Gemini CLI, and any other AI coding assistant that reads project-level context. A skill for "generate a commit message" activates when you ask for one; a skill for "code review" runs your team's actual checklist, not a generic pass. Spark hosts over 2,000 community-built and organization-published skills - scored, activity-tracked, and kept fresh.

When to use a Skill

Use a skill when you find yourself writing the same detailed instruction more than once. If you paste the same 200-word context about your project conventions into every Claude session, that's a skill waiting to be written. If your team has review standards that differ from Claude's defaults, a review skill enforces them every time.

Skills are also the right layer for integrations. A GitHub skill knows how to read your repo's PR template. A Linear skill understands your issue format. A Notion skill translates between Claude's output and Notion's block API. These aren't prompts - they're behavioral modules.

How to install a Skill

Most skills on Spark are a single file. Download it, drop it into .claude/skills/ in your project root (or the equivalent path your IDE recognizes), and you're done. No API keys, no server, no configuration step. Some skills include companion config files or example inputs; the detail page shows exactly what you get.

What to look for

The score on each skill reflects GitHub freshness, documentation quality, and community ratings. Skills above 70 are actively maintained. The activity dot - green for active, yellow for slow, orange for stale - shows upstream health. Sort by score to find community-validated picks. Sort by newest to see what shipped this week.