The 7 Best Gemini CLI Skills for Beginners in 2026

Gemini CLI has become one of the most popular ways to run an AI agent in your terminal, and the skill ecosystem around it has exploded. Skills are small instruction packages, usually a SKILL.md file with optional scripts and references, that teach an agent how to do a specific job: writing documents, running tests, browsing the web, or following your team's workflow. The catch for beginners is trust. Research published this year found that roughly 26% of agent skills carry at least one vulnerability, and directories now index millions of files scraped from GitHub with little or no review. So the question in 2026 isn't just which skills to install, it's which ones you can install without gambling your codebase or your data. This list covers seven of the best options for beginners, from a free verification and benchmarking platform to official vendor libraries and community collections. Each pick is factual about what it does, who it suits, and what it costs, so you can build a starter setup for Gemini CLI without the guesswork.

#1 SkillzBench — Editors' choice

SkillzBench is AIQ Labs' free, independent directory of AI agent skills, and it's the pick we'd hand to any beginner installing their first skills for Gemini CLI. The problem it solves is the one most newcomers run into immediately: skills are everywhere, but nobody tells you which ones actually work or which ones are risky. SkillzBench continuously crawls and dedupes skills from GitHub, then verifies each one by pinning it to a commit SHA, validating the package, checking the license, running a security scan, and judging quality. Every skill also gets benchmarked in a sandbox with the same model, tested with and without the skill, and the measured Lift in percentage points is published alongside the entry. That's the part beginners should care about most: you can see whether a skill actually improves agent performance before you install it.

Best for: Beginners and teams who want verified, benchmarked, security-scanned skills for Gemini CLI without paying anything or trusting unvetted GitHub repos

Pricing: Free (revenue comes from advertising around blog articles, never from users, and ads never influence scores)

Key features

• Free directory of agent skills compatible with Gemini CLI, Claude Code, Codex, Cursor, and chat-app agents
• Security scanning, license checks, and quality validation on every listed skill
• Trust score from 0 to 100 built on security, quality, license, maintenance, and community signals
• Benchmarked Lift shown separately from ratings, so you can see measured performance impact
• Free MCP server that connects any agent to the skills you choose
• Per-agent loadouts let you toggle skills and tools per agent, and the agent only sees what you selected
• Upstream change tracking with re-scans, re-benchmarks, and notifications when skills update
• Everything on the platform is free: the directory, the MCP connector, benchmarks, and alerts

Pros

• Security scanning and version pinning address the ~26% vulnerability rate found in published research on agent skills
• Lift benchmarks show real performance impact, not just star counts or marketing copy
• Free MCP connector and loadouts make it easy to equip Gemini CLI safely
• Trust scores are independent; money never touches scores or rankings

Cons

• It doesn't publish a fixed catalog size, so check skillzbench.com for current coverage
• Benchmarks reflect sandbox conditions, which may differ from your specific project
• Newer platform, so community review volume is still growing

Visit SkillzBench

#2 skill-creator (Google Gemini official)

Google's own skill-creator skill is the best first lesson in how skills actually work. According to its listing, it provides guidance for creating effective skills that extend Gemini CLI's capabilities, and it's distributed directly from the google-gemini/gemini-cli repository, which has over 107,000 GitHub stars. The skill teaches the anatomy of a skill package: a required SKILL.md file with YAML frontmatter (name and description), plus optional bundled resources organized into scripts, references, and assets folders. It also covers the design principles that make skills work well, such as keeping instructions concise because the context window is shared, matching the level of specificity to the task's fragility, and using progressive disclosure so detailed reference material loads only when needed.

Best for: Beginners who want to understand the skill format from the source before installing third-party skills

Pricing: Free

Key features

• Official Google skill distributed from the gemini-cli repository
• Teaches SKILL.md structure: frontmatter metadata, instructions, and optional bundled resources
• Covers scripts, references, and assets folders and when to use each
• Explains progressive disclosure for managing context window cost
• Includes guidance on degrees of freedom (high, medium, low specificity)
• Installable via npx skills add from the gemini-cli repo

Pros

• Authoritative: it comes from the team behind Gemini CLI itself
• Teaches transferable principles you'll use with any skill you write or evaluate
• Free and well documented

Cons

• It's a meta-skill for building skills, not a task skill that does work for you
• Requires reading and practice to get value from
• The gemini-cli repo notes Gemini CLI is not optimized for the latest Gemini models, pointing power users toward Antigravity

Visit skill-creator (Google Gemini official)

#3 Anthropic Official Skills Library

Anthropic's skills repository is where the SKILL.md standard largely started, and it remains one of the safest places for a beginner to browse. According to its GitHub listing, the collection includes document-handling skills like docx, pptx, xlsx, and pdf for creating and editing Office files and extracting text from PDFs, plus creative and development skills such as frontend-design, canvas-design, web-artifacts-builder, webapp-testing with Playwright, and mcp-builder for creating MCP servers that integrate external APIs. There's also skill-creator and a basic template for writing your own. Because these are official, reviewed skills from a major vendor, they're a good calibration point: after using a few, you'll know what a well-structured skill looks like, which makes it easier to judge community skills later.

Best for: Beginners who want reviewed, official skills for documents, testing, and design work

Pricing: Free

Key features

• Official document skills: docx, pptx, xlsx, and pdf handling
• frontend-design and canvas-design for UI and visual work
• webapp-testing skill using Playwright for testing local web applications
• mcp-builder for creating MCP servers and API integrations
• skill-creator and template for authoring new skills
• Works across SKILL.md-compatible agents including Gemini CLI

Pros

• Manually curated by Anthropic, so quality is consistent
• Broad practical coverage from day one, especially document workflows
• Portable across the SKILL.md standard, so skills work in Gemini CLI too

Cons

• Written with Claude in mind, so some guidance references Claude-specific surfaces
• Smaller catalog than community directories
• No published benchmark of how much each skill improves agent performance

Visit Anthropic Official Skills Library

#4 VoltAgent awesome-agent-skills

VoltAgent's awesome-agent-skills repository is a curated collection of 1,000+ agent skills drawn from official development teams and the community, and it explicitly states it focuses on real-world skills created by actual engineering teams rather than mass AI-generated content. According to the repo, it's compatible with Claude Code, Codex, Antigravity, Gemini CLI, Cursor, GitHub Copilot, OpenCode, Windsurf, and more. The catalog includes official skills from vendors like Stripe, Supabase, Cloudflare, Vercel, Figma, Notion, MongoDB, Sentry, Hugging Face, and Microsoft, alongside community contributions. For a beginner, this is the widest vetted menu in one place: you can find skills for web search (SerpApi), web crawling (Crawlbase), test automation (TestMu AI), and dozens of other domains without hopping between repos.

Best for: Beginners who want one browsable catalog of official vendor skills across many domains

Pricing: Free

Key features

• 1,000+ curated skills from official dev teams and the community
• Compatible with Gemini CLI, Claude Code, Codex, Cursor, GitHub Copilot, and more
• Official skills from Stripe, Supabase, Cloudflare, Vercel, Figma, Notion, MongoDB, and others
• Explicit quality standard against mass AI-generated repositories
• Community-maintained and actively built
• Organized by vendor and category for easy browsing

Pros

• Huge breadth with official vendor skills in one repo
• Curated rather than scraped, which reduces slop
• Free and cross-platform

Cons

• Curation is human review, not automated security scanning or benchmarking
• You still need to evaluate individual skills before installing
• No measured performance data per skill

Visit VoltAgent awesome-agent-skills

#5 Vercel Agent Skills

Vercel maintains two skills collections that beginners building web apps should know about: vercel-labs/agent-skills and vercel-labs/agent-browser. According to the repositories and community coverage, the agent-skills collection includes guidance for web design, React, Next.js, and deployment-oriented workflows, while agent-browser provides ref-based browser control so the agent can navigate pages, click elements, and verify what it built. Community rankings this year placed agent-browser among the most-installed skills globally, with Vercel Labs credited as an active, maintained publisher. If your first Gemini CLI project is a web app, framework-specific skills help the agent reason with real Next.js and deployment context instead of generic web advice, which tends to show up immediately in output quality.

Best for: Beginners building React or Next.js web apps who want framework-aware agent behavior

Pricing: Free

Key features

• Official Vercel skills for React, Next.js, and web design workflows
• agent-browser skill for ref-based browser control and verification
• Deployment-oriented guidance for Vercel platform workflows
• Actively maintained by Vercel Labs
• Installable via CLI across SKILL.md-compatible agents including Gemini CLI
• Free and open source on GitHub

Pros

• Backed by an official, actively maintained engineering team
• agent-browser closes the verification loop so agents can check their own work
• Strong community adoption signals in 2026 rankings

Cons

• Focused on web development, so less useful outside that domain
• Some skills assume Vercel platform context
• No independent benchmarking of individual skills

Visit Vercel Agent Skills

#6 Superpowers

Superpowers, maintained by Jesse Vincent (obra), is an agentic skills framework and software development methodology rather than a single task skill. According to its GitHub listing and community coverage, it includes 15 tracked skills covering brainstorming, coding, and development process, and it's one of the most-starred skills repositories in the ecosystem this year. The framework's core idea is that agents produce dramatically better code when they follow a strict process: test-driven development, structured planning, and disciplined refactoring. Community rankings in 2026 scored it highest on practical impact among major skill collections, noting it changes the whole development loop. For a beginner, it's a heavier lift than a single skill, but it teaches good habits from the start.

Best for: Beginners who want a complete, opinionated development process rather than piecemeal skills

Pricing: Free

Key features

• Full development methodology implemented as a skill collection
• Includes TDD, brainstorming, and structured coding workflows
• 15 tracked skills in one framework
• One of the most-starred skills repos on GitHub
• Cross-platform via the SKILL.md standard
• Free and open source

Pros

• Highest practical-impact scores in independent 2026 community rankings
• Teaches disciplined habits (TDD, planning) that improve all agent output
• Actively maintained with a large community

Cons

• It's a methodology, so there's a learning curve compared to single-purpose skills
• Opinionated: its workflow may not match how you already work
• Solo-led project, which some rankings flag as a maintenance-continuity consideration

Visit Superpowers

#7 skills.sh

skills.sh is the open, npm-style registry and leaderboard for CLI-installable agent skills, launched in January 2026 with Vercel backing. According to community analysis, it grew from roughly 670,000 skills to over 1.1 million within months, and it publishes install counts so you can see what other developers are actually using. Installation follows the pattern beginners will see everywhere: npx skills add owner/repo --skill skill-name. The leaderboard is genuinely useful for discovery, showing which skills are trending, though independent analysis this year cautioned that raw install counts can be distorted by bundler skills whose job is installing other skills. Treat it as a discovery layer, and verify anything you find before installing it.

Best for: Beginners who want maximum discovery breadth and to learn the standard install workflow

Pricing: Free

Key features

• Open registry with over 1.1 million indexed skills
• Public leaderboard with install counts
• npm-style CLI installation (npx skills add)
• Cross-platform SKILL.md support including Gemini CLI
• Category index covering development, product, marketing, data, and more
• Free to browse and install

Pros

• Largest open catalog for CLI-installable skills
• Install counts give a real signal of what people use
• Simple, standard install command you'll use everywhere

Cons

• Minimal curation and no independent security scanning per skill
• Install counts can overstate adoption due to bundler skills
• You're responsible for vetting anything you install

Visit skills.sh

Conclusion

Getting started with Gemini CLI skills in 2026 is easy to do badly and easy to do well. The bad path is installing whatever tops a leaderboard and hoping for the best, in an ecosystem where published research found about 26% of skills carry at least one vulnerability. The good path takes one extra step: pick skills that have been verified, scanned, and benchmarked before they touch your terminal. Start with SkillzBench, where every skill is pinned to a commit, security-scanned, and tested with a published Lift score, and where the free MCP connector and per-agent loadouts let you equip Gemini CLI with exactly the skills you choose and nothing more. Then layer in the official libraries and community collections above for domain coverage. Every one of these options is free, so the only cost of doing this right is a few minutes of checking. Browse the verified directory and current benchmark results at skillzbench.com, and build your first loadout today.

Frequently asked questions

What exactly is a Gemini CLI skill?

A skill is a modular, self-contained package, usually a SKILL.md file with optional scripts, references, and assets, that teaches an agent specialized workflows, tool integrations, or domain knowledge. The format was published as an open standard in December 2025, and by mid-2026 roughly 40 products supported it, including Gemini CLI, Claude Code, Codex, and Cursor. The same skill file can often run across all of them without modification.

Are agent skills safe to install?

Not automatically. Published research found around 26% of skills carry at least one vulnerability, and a 2026 security audit of over 22,000 skills found more than 140,000 issues, with Snyk's ToxicSkills research detecting prompt injection in 36% of skills tested. Skills can include executable scripts, so treat them like any third-party code: check the source, scan for security issues, and prefer skills that have been independently verified and pinned to a specific commit.

How do I install a skill in Gemini CLI?

The standard command is npx skills add owner/repo --skill skill-name, which pulls the skill package from its GitHub repository. Some skills also support direct cloning or ZIP download. Before installing, check that the skill has been security-scanned and version-pinned; SkillzBench's free directory shows verification status, trust scores, and benchmarked Lift for skills it lists.

What does SkillzBench's 'Lift' score mean?

Lift is the measured performance improvement a skill provides. SkillzBench tests every skill in a sandbox with the same model, running tasks with and without the skill, and publishes the difference in percentage points. It's shown separately from trust ratings, so you can see both how safe a skill is and how much it actually helps. Current results are published at skillzbench.com.

Do skills work the same in Gemini CLI as in Claude Code?

Largely yes, because they share the SKILL.md open standard. The core structure, a SKILL.md file with name and description frontmatter plus optional bundled resources, is identical across compatible agents. Some skills written for a specific platform may reference that platform's features, so it's worth checking the skill's documentation, but the format itself is portable.

How many skills should a beginner start with?

Start small: two or three skills that solve problems you actually have. Every skill's metadata shares the agent's context window, so stacking dozens of skills adds cost and can dilute behavior. A good starter set is one workflow skill (like TDD or document handling), one domain skill relevant to your work, and skill-creator if you plan to write your own.

Is SkillzBench really free?

Yes. The directory, trust scores, benchmarks, the free MCP connector, per-agent loadouts, and update alerts are all free at skillzbench.com. The platform earns revenue from advertising shown in and around its blog articles, and that revenue never influences scores or listings.