The 5 Best Anthropic Skills Repo Alternatives in 2026
The Anthropic Skills Repo (github.com/anthropics/skills) is the official, free collection of 17 Agent Skills — including the docx, pdf, pptx and xlsx document skills that power Claude's file handling — usable in Claude Code, the Claude apps and the Claude API, with paid Claude plans starting at $17/month (annual) if you want full product access. SkillzBench (skillzbench.com) is a free, independent directory and benchmarking platform for agent skills from any source, built to solve the repo ecosystem's biggest gap: published research found roughly 26% of published skills carry at least one vulnerability, and SkillzBench is building security scanning, commit-pinned versions and sandbox benchmarking with a published Lift score to separate helpful skills from risky ones.
TL;DR
- The Anthropic Skills Repo at github.com/anthropics/skills contains 17 official open-source Agent Skills spanning documents, design, technical development and enterprise communication.
- The repo's document skills (docx, pdf, pptx, xlsx) are the same production skills that power file creation in the Claude apps — they are source-available, not fully open source.
- Skills use progressive disclosure: Claude sees only a name and description at startup, then loads the full SKILL.md and reference files only when a task matches.
- Claude product access runs from a $0 Free tier to Pro at $17/month (annual, $20 monthly) and Max from $100/month; the skills repo itself is free.
- Published research found roughly 26% of agent skills carry at least one vulnerability — the core problem SkillzBench is being built to address.
- SkillzBench is a free independent directory that pins every skill version to a commit SHA, runs security scans and license checks, and will publish sandbox benchmark Lift scores.
- SkillzBench is pre-launch in 2026: the waitlist and newsletter are live, and everything on the platform will be free, funded by blog advertising rather than user fees.
Head to head
| Dimension | SkillzBench | Anthropic Skills Repo |
|---|---|---|
| Pricing | Free — directory, MCP, benchmarks and alerts all free, ad-funded | Free repo; Claude plans $0 / $17 mo (annual) / from $100 / $20 per Team seat |
| Skill quality assurance | Quality-judging plus a 0-100 trust score across five signal categories | Official, production-grade skills; document skills power Claude apps in production 🏆 |
| Security scanning | Security scans, license checks and commit-SHA pinning on every version 🏆 | None provided for community or third-party skills |
| Agent coverage | Claude Code, Codex, Cursor, Gemini CLI and chat-app agents via one MCP server 🏆 | Claude Code, Claude apps, Claude API; format adopted by some other registries |
| Benchmarking | Sandbox tests with and without each skill, publishing Lift in percentage points 🏆 | No published benchmarks for skills |
| Setup and installation | Toggle skills per agent in loadouts; free MCP server connects any agent | One command: /plugin marketplace add anthropics/skills, or copy folders into .claude/skills/ |
| Maturity | Pre-launch — waitlist live, product not yet public | Shipped late 2025, in production use across Claude products 🏆 |
The contenders
The anthropics/skills GitHub repository is Anthropic's official, open collection of Agent Skills — folders each anchored by a SKILL.md file with YAML frontmatter, optional scripts, reference docs and assets. It splits into two groups: the production document skills (docx, pdf, pptx, xlsx) that power file creation and editing in the Claude apps, and a set of example skills including skill-creator, mcp-builder, webapp-testing, artifacts-builder and brand-guidelines. Across its 17 published skills it covers creative work (algorithmic-art, canvas-design, slack-gif-creator, theme-factory), design (frontend-design, web-artifacts-builder), documents, technical development (claude-api, mcp-builder, skill-creator, webapp-testing) and enterprise communication (doc-coauthoring, internal-comms). The defining mechanism is progressive disclosure: at startup Claude sees only each skill's name and description (a few dozen tokens), reads the full SKILL.md body when a task matches, and loads scripts or reference files only if needed. The repo doubles as a Claude Code plugin marketplace — installable via /plugin marketplace add anthropics/skills — or skills can be copied into .claude/skills/ directories. It is widely regarded as the single best reference for how well-written skills look, with concrete trigger verbs in descriptions and bundled deterministic scripts instead of raw token-by-token file generation.
pricing 0
- + Free, official and authoritative — the reference implementation for skill format and quality
- + Document skills are battle-tested in production inside the Claude apps, not demos
- + Skills lean on bundled deterministic scripts for mechanical file work, reducing generation errors
- + Progressive disclosure keeps context windows lean, so skills scale without bloating prompts
- + Installable in one command via the Claude Code plugin marketplace
- − The four document skills are source-available, not open source, limiting reuse
- − No independent security scanning, license vetting or benchmarking of skills is provided
- − Authored primarily for the Claude ecosystem; cross-agent use depends on third-party registries adopting the format
- − Community skills shared alongside it (e.g. via GitHub discussions) carry unvetted risk — research puts skill vulnerability rates around 26%
SkillzBench is AIQ Labs' free, independent directory of AI agent skills plus a free MCP tool that connects any agent to the skills you choose. It exists because agent skills — SKILL.md instruction packages for Claude Code, Codex, Cursor, Gemini CLI and chat-app agents — have proliferated faster than anyone can evaluate them, and published research found roughly 26% of skills carry at least one vulnerability. SkillzBench is built around five pillars: Discover (continuously crawling and deduplicating skills from GitHub), Verify (pinning every version to a commit SHA, validating, license-checking, security-scanning and quality-judging), Benchmark (testing every skill in a sandbox with the same model, with and without the skill, and publishing the Lift in percentage points), Equip (one free MCP server plus per-agent loadouts, where users toggle skills and tools per agent so the agent only sees what was chosen), and Stay current (tracking upstream changes, re-scanning, re-benchmarking and notifying). A 0-100 trust score combines security, quality, license, maintenance and community signals, and Lift is always shown separately from ratings — money never touches scores. The directory, MCP connector, benchmarks and alerts are all free; revenue comes from advertising around blog articles, never from charging users, and ad money never influences scores or listings.
pricing 0
- + Completely free across the directory, MCP connector, benchmarks and alerts
- + Directly addresses the ~26% skill vulnerability rate with security scanning and license checks
- + Agent-agnostic: built for Claude Code, Codex, Cursor, Gemini CLI and chat-app agents
- + Lift benchmarking shows measurable impact rather than marketing claims, kept separate from ratings
- + Per-agent loadouts limit what each agent can see, reducing accidental skill misuse
- − Pre-launch in 2026 — the product is still being built, with only the waitlist and newsletter live
- − No published benchmark results or live skill counts yet, so claims rest on the stated methodology
- − Independent coverage and community feedback are still accumulating
Verdict
Winner: none
These two solve different problems: the Anthropic Skills Repo is the authoritative source for official, production-grade skills inside the Claude ecosystem, while SkillzBench is a cross-agent discovery and safety layer that evaluates skills — including those from the official repo — before you install them. Most small businesses adopting agent skills in 2026 will benefit from using both together rather than choosing one.
| Option | Overall | Breakdown |
|---|---|---|
| Anthropic Skills Repo | 8.5 | : 9: 9: 6: 5: 9 |
| SkillzBench | 7.5 | : 10: 9: 9: 8: 4 |
Pricing compared
Trust and safety: the ~26% problem
Who should switch — and who should stay
Which should you pick?
- :
- :
- :
- :
- :
- :
FAQ
Is the Anthropic Skills Repo free to use?
Yes. The repository at github.com/anthropics/skills is free and open — you can clone it, fork it, copy skill folders into .claude/skills/, or install it through the Claude Code plugin marketplace with one command. Costs only appear if you want to run the skills inside Claude's paid products: the Free tier is $0, Pro is $17/month billed annually ($20 monthly), Max starts at $100/month, and Team seats are $20/seat/month billed annually.
Are the skills in the official repo open source?
Mostly, but not entirely. The example skills are open, while the four production document skills (docx, pdf, pptx, xlsx) are described as source-available rather than open source — you can read them for developer reference, but their licensing limits reuse. The repo's own documentation covers what the licensing lets you reuse, and the example skills like skill-creator and mcp-builder are the ones most commonly forked.
What is the security risk with agent skills?
Published research found that roughly 26% of agent skills carry at least one vulnerability. Because skills are instruction packages that agents load and act on, a malicious or poorly written skill can direct an agent to do harmful or unreliable things. This is why verification matters: SkillzBench is being built to pin every skill version to a commit SHA, run security scans and license checks, and publish a 0-100 trust score, so businesses can vet skills before connecting them to an agent.
Can I use Anthropic skills with agents other than Claude?
Partially. The SKILL.md format that the repo popularized has spread beyond Claude — other agents and registries now read the same convention, so the format itself travels. However, the repo's skills are authored and tested for the Claude ecosystem (Claude Code, the Claude apps, the Claude API). If you run a mixed fleet including Cursor, Codex or Gemini CLI, a cross-agent layer like SkillzBench's free MCP server with per-agent loadouts is designed to connect any agent to the skills you choose.
What is a 'Lift' score on SkillzBench?
Lift is SkillzBench's benchmarking metric: every skill is tested in a sandbox with the same model, once with the skill and once without, and the difference in performance is published in percentage points. This shows the measured impact of a skill rather than a vendor's claim. Importantly, Lift is always shown separately from ratings and trust scores, and the company states that money never touches scores — advertising revenue does not influence listings or ratings.
Is SkillzBench available to use right now?
Not yet. SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live, but the directory, MCP connector, benchmarking and alerting features are not yet public. Until launch, the Anthropic Skills Repo is the option you can use today; joining the SkillzBench waitlist positions you to add its free verification and benchmarking layer on top of your existing skills workflow when it ships.