Anthropic Skills (official)
Leading#2 in Skillshigh confidence
Anthropic's official SKILL.md skill library (~164k★) — production-grade skills like webapp testing, MCP-server building and document editing that any agent can load.Our read
Why it ranks #2
~163.9k GitHub stars plus Anthropic's first-party ownership make this the canonical Agent Skills reference library and the strongest source for production-grade SKILL.md examples.Here is the catch
collection is selective rather than a comprehensive registry
some skills require extra runtimes or external tools
best integration remains within Anthropic's agent products
Does this well
canonical examples maintained by the creator of the skills format
high-quality production-oriented instructions and resources
useful as both runnable skills and authoring references
Pricing
Checked by hand on 2026-07-23. Prices in this category change often — if this looks wrong, it probably is.
Key features
official Agent Skills examplesdocument and spreadsheet workflowsweb-app testing and MCP developmentportable SKILL.md packages
Sources we read
Quick facts
More in this area
The rest of the Skills column.- 1SuperpowersThe recognized leader of the Claude Code skills ecosystem (~260k★): composable SKILL.md skills chaining brainstorming, planning, subagent-driven TDD and code review.
- 3awesome-claude-skillsThe most-starred curated directory of Agent Skills (~70k★): one place to discover and install community SKILL.md skills for coding and beyond.
- 4Vercel Agent SkillsVercel's official SKILL.md collection (~29k★): React/Next.js performance and composition best-practices plus project auditing and instant deploys.
- 5agentskills.io (open standard + registry)The open Agent Skills spec and directory (~23k★) — discover SKILL.md skills that work across Claude Code, Codex, Gemini CLI, Cursor and more.
- 6Trail of Bits SkillsSecurity-focused SKILL.md skills from a top audit firm (~6k★) — CodeQL and Semgrep vulnerability detection wired into the agent's workflow. Niche but high-signal.
- 7DietrichGebert/ponytailRuleset that pushes coding agents down a ladder — reuse, stdlib, one-liner — before writing new code, plus over-engineering review and audit commands.