The Best skills.sh Alternative for Teams in 2026
skills.sh is Vercel's free, open directory for AI agent skills: browse a leaderboard, run npx skills add owner/repo, and the same skill installs into a wide range of coding agents — sources report agent support anywhere from 18 to 70+ depending on the tool and date. It's excellent for discovering and trying popular public skills, but it's a discovery surface first: everything is public, installs point at community sources with no publish/rollback step, and there's no review gate or team access control. If that's all you need, stay put. If your team needs private skills, versioning, review workflows or independent quality/security verification before install, the strongest alternatives are localskills.sh (a full private registry with team controls), a plain GitHub repo (zero infrastructure), Agentman's Agent Skills (hosted library from $30/mo), ClawHub (curated, versioned), and SkillzBench — a free platform being built to independently verify, security-scan and benchmark skills before you ever install one.
Head to head
| Dimension |
|---|
The contenders
skills.sh is Vercel's open directory and leaderboard for AI agent skills, launched in January 2026. It indexes community skills from GitHub and installs them with one command (npx skills add owner/repo) into a broad set of coding agents — sources report agent support ranging from 18 to 70+ tools, including Claude Code, Cursor, Codex, GitHub Copilot and Windsurf. Skills are simple SKILL.md folders with optional scripts and templates, and the CLI auto-detects which agent you're using and drops the skill into the right location. The directory is free with no sign-in required to install, the CLI is MIT-licensed, and indexed skills are scanned by partner auditors (Gen Agent Trust Hub, Socket, Snyk) with failing skills excluded. Its leaderboard ranks by install count — the top skill reads 3.6M installs — and it has grown Packs, Topics, Official and Audits sections. The core limitation, per multiple reviews: it's a discovery surface, not a team registry — everything is public, there's no publish/rollback step, and popularity doesn't prove quality or security.
pricing 0
- + Free with no sign-in needed to install; CLI is MIT-licensed open source
- + Same skill installs into many agents — sources report 18 to 70+ supported tools
- + Strong adoption signal: 8M+ total installs and a top skill at 3.6M installs
- + Indexed skills are scanned by partner auditors and failing skills are excluded
- − Everything is public — no home for private internal skills
- − No publish step that freezes a version, so no meaningful rollback or diff between reviewed and installed
- − No review gate, team roles, access control or audit trail
- − Popularity ranking doesn't prove correctness, maintenance quality or security
localskills.sh is a full agent skills registry rather than a directory, and the research positions it as the strongest skills.sh replacement for teams. You publish skills as versioned packages and one published skill installs into Claude Code, Cursor, Windsurf, GitHub Copilot, Cline, Codex CLI, Aider and more — the CLI writes each tool-native format from one source of truth. Every skill is a real package (root SKILL.md, multi-file folders up to 100 MB and 500 files per version) with versioning and rollback, plus public, private and unlisted visibility and built-in download analytics. Org structure includes folders, namespaces, teams, member roles and custom roles, and folder-level access restrictions. Enterprise plumbing covers SAML SSO (Okta, Cloudflare Access, Authentik templates), SCIM sync, audit logs and verified-domain auto-join. Bidirectional GitHub sync lets review happen in a normal pull request, and an MCP server with OAuth lets agents search and install skills without a human in the loop. Its honest limit: it's a registry, so it asks you to publish rather than just point at a repo.
pricing See pricing
- + Real private distribution with versioning and rollback — the core thing skills.sh lacks
- + Multi-tool installs from one source of truth, including tool-native formats
- + Full enterprise plumbing: SSO, SCIM, audit logs
- + GitHub PR-based review workflow
- − Requires a publish step — overhead if you never need private distribution or team controls
- − No published benchmark of whether skills actually improve agent performance
SkillzBench is AIQ Labs' free, independent platform for discovering, verifying and benchmarking AI agent skills, plus a free MCP tool that connects any agent to the skills you choose. Its premise: skills are everywhere, but nobody can tell which ones actually help — published research found ~26% of skills carry at least one vulnerability. SkillzBench is being built around five pillars: Discover (continuously crawl and dedupe skills from GitHub), Verify (pin every version to a commit SHA, validate, license-check, security-scan and quality-judge), Benchmark (test every skill in a sandbox with the same model, with and without the skill, and publish the Lift in percentage points), Equip (one free MCP server plus per-agent loadouts — toggle skills per agent so the agent only sees what was chosen), and Stay current (track upstream changes, re-scan, re-benchmark, notify). A 0-100 trust score combines security, quality, license, maintenance and community signals; Lift is always shown separately, and money never touches scores. Everything is free — the directory, MCP connector, benchmarks and alerts — with revenue from blog advertising only. SkillzBench is pre-launch and actively being built; the waitlist is live.
pricing 0
- + Free across the board, funded by advertising — money never touches scores
- + Independent verification and security scanning before you install anything
- + Benchmark Lift published separately from ratings, so you see real performance impact
- + Per-agent loadouts via a free MCP connector
- − Pre-launch — the product is not yet public; only the waitlist/newsletter is live
- − Not a private distribution registry — it verifies and equips, it doesn't host your team's private skills
The zero-infrastructure alternative: Anthropic maintains an official skills repo, and community 'awesome' lists function as a de facto skills directory in README form. The workflow is simple — clone or copy the skill folder into .claude/skills/ and your agent has it. Per the research, the strengths are real: it's free, fully transparent, easy to fork, and review happens naturally because you read the file before copying it. That makes it the safest trust model of any option here — your team reviews exactly what it installs, in a normal pull request, with git history as your audit trail. The limits are equally clear: installs are manual, there's no multi-tool translation (a Claude-format skill doesn't automatically become a Cursor-format one), no versioning or rollback beyond what git gives you, no analytics, and no discovery beyond what the README lists. For a small team comfortable with git, it's often the right call; for anyone else, the manual workflow is the cost.
pricing 0
- + Free and fully transparent
- + Review happens naturally — you read the file before copying it
- + Zero new infrastructure
- + Easy to fork and customize
- − Installs are manual
- − No multi-tool translation between agent formats
- − No versioning, rollback, analytics or discovery tooling
Agentman's Agent Skills is a hosted, private skills library aimed at teams — the research's clearest example of a commercial registry with per-seat pricing. The free plan gives up to 10 private skills and 3 members with no credit card, including version history and cloning from the public library, usable from Claude, ChatGPT and Cursor. Team ($10/seat/mo, 3-seat minimum, so $30/mo entry) adds unlimited members and private skills with a 14-day trial. Business ($25/seat/mo, 5-seat minimum, $125/mo entry) adds per-skill access control, use-only sharing (run a skill without seeing its instructions), approvals before changes publish, and access/activity visibility. Enterprise offers private and on-premise deployment, and a Partner track gives agencies multi-tenant workspaces per client. Everything reads one shared library over a single MCP connection, with semantic search and .skill package import/export. For a small business that wants private skills with review gates and doesn't want to self-host, it's the most turnkey paid option in this comparison.
pricing 30
- + Genuinely free starter plan with no credit card
- + Turnkey private library with review gates and approvals
- + Use-only sharing is a real security control
- + Clear, published per-seat pricing
- − Seat minimums push the real entry cost to $30/mo (Team) or $125/mo (Business)
- − Curated catalog of 90 production skills is small compared with open directories
ClawHub grew out of the OpenClaw ecosystem and is the research's 'quality over quantity' alternative: roughly 3,200 skills with 1.5M+ downloads, the smallest major catalog by far but the most curated. It has its own CLI (npx clawhub@latest install skill-slug), vector-powered semantic search, and — unlike skills.sh — proper versioning with rollback support. OpenClaw creator Peter Steinberger built it explicitly as the curated counterpoint to volume-led catalogs. The trade-off is agent reach: it supports OpenClaw plus bridges rather than the wide native agent support of skills.sh, so teams not on OpenClaw should check bridge compatibility before committing. For small businesses that want a smaller, versioned, curated catalog with rollback instead of an 80K+ skill firehose ranked by popularity, ClawHub is the research's clearest curated pick.
pricing 0
- + Most curated of the major catalogs — quality over quantity by design
- + Versioning with rollback, which skills.sh lacks
- + Semantic search across the catalog
- − Smallest major catalog (~3,200 skills)
- − Agent support is OpenClaw plus bridges — limited native reach
SkillsMP is the volume leader among agent skills marketplaces: the research reports it topping 351,000 skills by early March 2026, with a later figure of ~1.9 million public skills scraped from GitHub. It's a community-run discovery platform that crawls GitHub for SKILL.md files and indexes them with AI-powered semantic search, with a deep category taxonomy (89K tools skills, 70K development skills, 60K business skills). It supports Claude Code, Codex CLI and ChatGPT — three agents, far fewer than skills.sh. Crucially, the research is blunt about its quality control: none beyond GitHub stars and date, so you inspect before installing. For small businesses it works as a broad discovery surface — the semantic search is genuinely useful for finding niche skills in a giant catalog — but it offers no curation, no versioning and no security review, making it the option where the trust burden on your team is highest.
pricing 0
- + Largest catalog of any option — best raw discovery
- + Semantic search makes a giant catalog navigable
- + Free to browse
- − No quality gate beyond GitHub stars and date
- − No security review — inspect everything before installing
- − Supports only 3 agents
Anthropic launched Agent Skills (the feature most users know as Claude Skills) in October 2025 and published the spec as an open standard at agentskills.io on December 18, 2025. Inside Claude Code, Anthropic ships a small, manually curated directory and maintains a partner directory with skills from Atlassian, Canva, Cloudflare, Figma, Notion, Ramp, Sentry, Stripe and Zapier — everything there is reviewed. As of June 2026, roughly 40 skills-compatible products appear on the official showcase, including OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI and VS Code. For a small business, this is the highest-trust catalog in the research: manually curated, verified, from named vendors. The trade-off is breadth — the catalog is small compared with any community directory, and it's Anthropic-centric. If your team lives in Claude Code and wants only vendor-reviewed skills, start here; for anything beyond that, pair it with a verification platform like SkillzBench.
pricing 0
- + Highest trust model — everything reviewed and verified
- + Skills from established, named software vendors
- + Anchors an open cross-vendor standard
- − Small catalog compared with any community directory
- − Anthropic/Claude-centric
Verdict
Winner: localskills.sh
For small-business teams, the research is clear that the best skills.sh alternative depends on the gap you're closing. If the gap is private distribution, versioning and team review, localskills.sh is the strongest replacement — it's a full registry with rollback, roles, SSO and multi-tool installs. If the gap is zero cost and zero infrastructure, a GitHub repo beats everything. If the gap is trust — knowing which skills actually help and which are risky — no directory answers it, because every catalog ranks by popularity, not verified quality. SkillzBench is being built to close exactly that gap for free: security scanning, commit-pinned versions, sandbox benchmarks with published Lift, and a free MCP connector. It's pre-launch, so teams needing private distribution today should pick localskills.sh — but for verification and benchmarking, nothing else in this comparison offers it.