The Best Source-Verified Claude Code Skills (2026)
A curated list of real Agent Skills sourced from Anthropic's official repository and the Composio community collection.
Most skill lists are filler. This one isn’t.
Every entry below exists in a real public repo with a real SKILL.md you can read before installing. That’s the only criteria that mattered when putting this together. Provenance first, not volume.
Official Anthropic skills
1. Webapp Testing
One of the most practical skills in Anthropic’s repo. It gives Claude a structured approach to testing and QA — not just “write some tests,” but a repeatable framework for test planning, edge case identification, and structured reporting.
See: Webapp Testing
2. MCP Builder
If you’re building around the Model Context Protocol, this skill makes Claude think more carefully about server design, capability boundaries, and protocol compliance. Useful whether you’re writing an MCP server from scratch or reviewing someone else’s.
See: MCP Builder
3. Claude API
Anthropic’s API-oriented skill. A solid starting point for people building on the Claude developer platform — structured thinking around authentication patterns, request formatting, and error handling.
See: Claude API
4. Frontend Design
One of the stronger design-oriented official examples. Worth studying even if you mainly want to understand how well-structured skills are written — the way it applies design principles consistently rather than one-off aesthetic calls is a good pattern to borrow.
See: Frontend Design
5. Skill Creator
If you want to write your own skills, start here. It’s Anthropic’s meta-skill — a skill about writing skills. The format it uses is exactly what agents expect, and reading it will save you from most first-skill mistakes.
See: Skill Creator
Community skills from Composio
6. Content Research Writer
One of the clearest writing-oriented skills in Composio’s repo. It structures the research phase before writing begins — so Claude isn’t just generating from nothing, it’s working from actual inputs you’ve assembled together.
7. Lead Research Assistant
Structured prospect research before outreach or partnership conversations. It applies a consistent methodology across organizations, so you get comparable outputs rather than freeform notes that change shape every time.
8. Meeting Insights Analyzer
Better than a summarizer. It distinguishes between decisions made, action items assigned, and open questions — rather than collapsing everything into a bullet list that nobody reads. Genuinely useful for any team with recurring meetings.
See: Meeting Insights Analyzer
How to evaluate any skill before installing
A description can be misleading. A SKILL.md can’t.
Click the raw file link on any skill page and read the actual process. If it’s vague — “be a thoughtful engineer,” “apply best practices” — the output will be vague too. What you want to see: specific inputs named, a step-by-step process, a defined output format.
Also check when the upstream repo was last active. A skill in an actively maintained repo is more likely to reflect current tooling than one that’s been sitting untouched for two years.
How to install any of these
Every skill page shows the install command. The format:
npx skills add {owner}/{repo} --skill {path}
Or browse the full directory by category or audience if you’re looking for something more specific than what’s on this list.
📬 Weekly digest
Get the best new skills every Tuesday
3–5 hand-picked skills. Free forever.
More guides
March 17, 2026 · 3 min read
Why findskills.co focuses on source-verified skills
A quick explanation of why every listed skill now points to a real upstream GitHub folder and raw SKILL.md file.
March 16, 2026 · 5 min read
The complete guide to Agent Skills in 2026
What Agent Skills are, why they matter, how to install them, and how to choose the right ones for your workflow.
March 14, 2026 · 4 min read
How to build a skill stack for your workflow
Pick the right combination of Agent Skills for the kind of work you do instead of installing random individual skills.