The 5 Best SkillsMP Alternatives in 2026

SkillsMP calls itself an Agent Skills Marketplace, but functionally it is a large discovery index for public SKILL.md files sourced from GitHub. Its homepage reports 3,601,307 collected SKILL.md files, and you can browse by keyword, creator, repository, category (Tools, Business, Development, and others), or occupation using the U.S. Standard Occupational Classification system covering 23 major groups and 867 detailed occupations. Browsing and searching are free, with a free tier that allows one metadata export task per day at up to 500 rows and 100 API/MCP searches per day. The site is candid about its limits: it indexes references but does not certify or install them, and its terms state it does not endorse or verify the quality, safety, or functionality of any skill. Treat it as a map for discovery, then audit the source yourself before installing anything.

pricing 0/month

• + Free to browse and search, with a genuinely large catalog of millions of collected SKILL.md files.
• + Multiple discovery routes: keyword search, creator pages, repository collections, categories, and occupation-based browsing.
• + Free tier includes API and MCP search access, not just a web UI.
• + Cards link to public GitHub sources so you can inspect the code before installing.
• + The site is honest that it is a discovery index, not a certification, which sets correct expectations.

• − Its terms explicitly disclaim any guarantee of accuracy, reliability, or safety of indexed skills.
• − No security scanning, license checking, or benchmarking of listed skills.
• − Counts come from a fixed catalog snapshot, so individual cards can lag behind upstream repos.
• − Publication or indexing does not indicate real-world adoption or quality, as the site itself notes.
• − Free tier has tight export limits (one task, 500 rows per day) and shared capacity protection.

https://skillsmp.com

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. Where SkillsMP stops at indexing, SkillzBench continues: it continuously crawls and dedupes skills from GitHub, pins every version to a commit SHA, validates them, checks licenses, security-scans them, and quality-judges them. Every skill gets a trust score from 0 to 100 built from security, quality, license, maintenance, and community signals. It also benchmarks each skill in a sandbox, running the same model with and without the skill, and publishes the Lift in percentage points. Lift is always shown separately from ratings, and money never touches scores. Per-agent loadouts let you toggle skills and tools per agent so each agent only sees what you chose, and the platform tracks upstream changes, re-scans, re-benchmarks, and notifies you when things move. Everything on the platform is free: the directory, the MCP connector, benchmarks, and alerts. Revenue comes from advertising around blog articles, never from charging users, and ad money never influences scores or listings.

pricing 0/month

• + Everything is free: the directory, the MCP connector, benchmarks, and alerts.
• + Verification is the default, not an afterthought: version pinning, license checks, and security scans on every skill.
• + Benchmarks measure actual Lift in a sandbox with the same model, with and without the skill, so you see what a skill really does.
• + Lift is always shown separately from ratings and money never touches scores.
• + Per-agent loadouts give you control over exactly what each agent can see and use.
• + Upstream change tracking with re-scanning and notifications keeps loadouts current.

• − A curated, deduped directory means a smaller catalog than bulk indexes like SkillsMP's millions of collected files.
• − Verification and benchmarking take time, so brand-new skills may not appear instantly.
• − Advertising around blog articles funds the platform, which some readers may notice while browsing.

https://skillzbench.com

The most direct alternative to a third-party marketplace is the source itself. Anthropic, the company behind the SKILL.md format used by Claude Code, maintains a public skills repository on GitHub that includes widely used skills like skill-creator (which creates, edits, and benchmarks skills, with 181,414 stars and 21,482 forks on its repo) and frontend-design (guidance for distinctive, intentional visual design when building or reshaping UI). Because these come from the company that defines the format, compatibility and maintenance are about as trustworthy as it gets in this ecosystem. The tradeoff is scope: this is a small curated set, not a catalog, and there is no search interface, occupation browsing, or install tooling beyond what GitHub offers. You will want to pair it with a directory if you need broader discovery. Note that running these skills in Claude Code requires a Claude plan: Pro is $20/month (or $17/month billed annually), with Max tiers at $100 and $200 per month, per third-party pricing analysis.

pricing 0/month

• + First-party skills from the company that defines the SKILL.md format.
• + Free and open source, with strong community traction (181,414 stars on the skill-creator repo).
• + skill-creator doubles as a tool for building and evaluating your own skills.
• + No middleman: you read the actual source files directly on GitHub.

• − Small curated set; no way to discover skills from other creators.
• − No search, category, or occupation browsing; you need to know what you are looking for.
• − Manual installation; no MCP connector or loadout management.
• − Running the skills in Claude Code requires a paid Claude plan, with no reliable free CLI tier as of September 2026 per third-party analysis.

https://github.com/anthropics/skills

Skilltype often comes up in 'skills platform' searches, but be clear about what it is: a workplace skills-training platform for organizations, not a directory of AI agent skills. Its pricing page describes a central index of 25,000+ trainings, courses, and events, a skills ontology with gap analysis, individual learning paths, and manager dashboards. Pricing is a flat annual rate per employee band: $1,500 per year for 5 to 25 employees (about $60 per employee per year), $4,000 for 26 to 100, $9,000 for 101 to 500, and $18,000 for 501 to 1,500. Annual billing saves 20% versus monthly, and multi-year commitments save 10% at two years or 15% at three. Every plan includes SSO/SAML and API access, and the vendor says most teams are up and running in a day or two. It is a legitimate alternative if your actual problem is employee skill development, but it will not help you equip a Claude Code or Codex agent with SKILL.md packages.

pricing 1500/year

• + Flat annual rate per employee band, so adding a few people never bumps your bill.
• + Per-employee cost drops as you scale, from about $60/employee/yr down to about $12 at the Enterprise band.
• + Annual billing saves 20%, with additional multi-year discounts of 10-15%.
• + Vendor says setup takes a day or two with no professional-services team.

• − It is a human skills-training platform, not an AI agent skills directory; it does not serve the SKILL.md use case at all.
• − No free tier; the entry point is $1,500 per year.
• − Not viable for solo users or very small teams under 5 employees.
• − White-label agreements are priced individually and require a sales conversation.

https://www.skilltype.com

Vercel Labs publishes agent skills as free public GitHub repositories, and they show up prominently in SkillsMP's own popular picks. The standout is vercel-react-best-practices, which packages React and Next.js performance optimization guidelines from Vercel Engineering, triggered whenever you write, review, or refactor React or Next.js code. Vercel Labs also maintains agent-browser, a browser automation CLI for AI agents that handles navigation, form filling, screenshots, scraping, and testing, and can even automate Electron desktop apps like VS Code, Slack, Discord, and Figma. These are first-party skills from a single engineering team, which keeps quality high and scope narrow. There is no marketplace layer: you install from GitHub yourself, and the catalog will not grow beyond what Vercel publishes. It works best as a supplement to a broader directory rather than a replacement for one.

pricing 0/one_time

• + Free and first-party, written by the engineering team behind Next.js.
• + Highly practical for React/Next.js shops: performance guidance triggers exactly when reviewing that code.
• + agent-browser covers a genuinely wide range of browser and desktop automation tasks.

• − Narrow catalog limited to what Vercel Labs publishes.
• − No directory, search, or install tooling; everything is manual via GitHub.
• − No independent security scanning or benchmarking of the skills.

https://github.com/vercel-labs/agent-skills

Jesse Vincent's Superpowers repository is one of the most referenced skill collections in the ecosystem, with 15 indexed skills on SkillsMP covering planning, testing, debugging, and reviewing software changes. Its most popular skill, brainstorming, is described as a must-run before any creative work, whether building features, modifying behavior, or planning anything new, and it shows 300.0k in popularity signals on SkillsMP's card. The collection is aimed squarely at software engineering workflows: sharpening plans, developing software, and reviewing changes. Like other single-maintainer repositories, it is free and the maintainer is well known in the community, but there is no verification layer, no benchmarking, and no install tooling beyond GitHub itself. It is a strong supplement for engineering teams, and pairs well with a directory that can tell you whether these skills actually help.

pricing 0/one_time

• + Free, with a well-known maintainer and strong community traction (300.0k popularity signal on the brainstorming card).
• + Covers a coherent engineering loop: plan, test, debug, review.
• + Skills are composable, so you can adopt the pieces that fit your workflow.

• − Only 15 indexed skills; strictly an engineering-focused collection.
• − No security scanning or independent benchmarking.
• − Manual GitHub installation with no directory or loadout tooling.

https://github.com/obra/superpowers