Sign inGet started
Home
AI Power Users

The 9 Best Claude Skills for Writing in 2026

Discover the 9 best Claude Skills for writing in 2026. Match any house style or client voice and level up your drafts. See the full ranked list now.


If you write for a living, you've probably noticed that Claude is already a decent drafting partner. What it lacks out of the box is the specific procedural knowledge that makes writing sound like you, follow a house style, or match a client's voice. That's what Claude Skills are for. A skill is a folder with a SKILL.md file that packages instructions, examples, and sometimes scripts, and the agent loads it only when a task matches its description. According to Anthropic's engineering blog, skills use progressive disclosure so the agent reads just the name and description at startup, then pulls the full instructions when they're relevant. That means you can keep a whole shelf of writing workflows on hand without bloating every conversation. The catch is that anyone can publish a skill, and research from Snyk found prompt injection in a large share of skills tested. So the real question in 2026 isn't whether writing skills exist. It's which ones are worth trusting. This list covers nine options, from a free security-first directory to curated GitHub collections and official Anthropic packages.

#1 SkillzBench — Editors' choice

SkillzBench is a free, independent directory of AI agent skills plus a free MCP connector that plugs any agent into the skills you choose. It exists because the skills ecosystem has a trust problem: published research found that roughly 26% of skills carry at least one vulnerability. SkillzBench's answer is to verify before you install. Every skill is pinned to a commit SHA, validated, license-checked, security-scanned, and quality-judged. Then it gets benchmarked in a sandbox with the same model, with and without the skill, and the difference is published as a Lift percentage. For writing skills, that matters because a prompt-injection payload hidden in a "copywriting" skill can quietly exfiltrate your drafts or your client data. SkillzBench shows you the Trust score from security, quality, license, maintenance and community signals, and it keeps Lift separate from ratings and money. Everything on the platform is free, including the directory, the MCP connector, benchmarks and alerts. The company earns revenue from advertising around its blog articles, never from charging users, and ad money never touches scores or listings. You can browse hundreds of writing, editing, and documentation skills, toggle the ones you want per agent through loadouts, and let your agent see only what you chose.

Best for: Writers, editors, and content teams who want to install writing skills without gambling on unvetted GitHub repos, and developers building agent workflows that need a trustworthy skills layer.

Pricing: Free

Key features

  • Free independent directory of agent skills, continuously crawled and deduped from GitHub
  • Every skill version pinned to a commit SHA, validated, license-checked, security-scanned and quality-judged
  • Benchmarks run in a sandbox with the same model, with and without the skill, published as Lift in percentage points
  • Trust score from 0 to 100 based on security, quality, license, maintenance and community signals
  • Free MCP server that connects any agent to the skills you select
  • Per-agent loadouts so each agent only sees the skills and tools you toggle on
  • Tracks upstream changes and re-scans, re-benchmarks and notifies you when a skill changes
  • Everything is free; revenue comes from blog advertising and never influences scores or listings

Pros

  • Security scanning and commit pinning address the real supply-chain risk in the skills ecosystem
  • Lift benchmarks show whether a writing skill actually improves output rather than just sounding plausible
  • The MCP connector and loadouts let you control exactly which skills an agent can see
  • No cost for the directory, benchmarks, or connector, and ad money stays out of the rankings

Cons

  • It's a directory and benchmarking layer, not a writing tool itself, so you still bring your own agent
  • The catalog is large and growing, so you'll want to filter by category and Trust score to find writing skills quickly

Visit SkillzBench

#2 Anthropic Official Skills

Anthropic's own skills repository is the reference implementation for the format and the safest starting point for writing-adjacent work. According to the VoltAgent curated collection, the official set includes docx for creating, editing and analyzing Word documents, doc-coauthoring for collaborative document editing, pptx for presentations, xlsx for spreadsheets, pdf for extracting text and handling forms, and internal-comms for status reports, newsletters and FAQs. There's also a skill-creator skill that guides you through building your own skills, and a brand-guidelines skill that applies Anthropic's brand colors and typography to artifacts. For writers, the doc-coauthoring and internal-comms skills are the most directly useful: they encode a repeatable process for turning rough notes into structured documents. The docx and pdf skills handle the mechanical side of getting words into the right file format. These are first-party skills, so they carry Anthropic's own maintenance and review. They're not a directory of community writing skills, and they won't teach Claude your personal voice, but they're a solid baseline before you add anything third-party.

Best for: Writers who want a trusted, first-party baseline for document creation and internal communication before adding community skills.

Pricing: Free (requires a Claude plan for use)

Key features

  • Official docx skill for creating, editing and analyzing Word documents
  • doc-coauthoring skill for collaborative document editing and co-authoring
  • internal-comms skill for status reports, newsletters and FAQs
  • pptx, xlsx and pdf skills for presentations, spreadsheets and PDF form handling
  • skill-creator skill that guides creation of new skills
  • brand-guidelines skill for applying consistent visual identity to artifacts
  • Maintained by Anthropic as the reference implementation of the Agent Skills format

Pros

  • First-party maintenance and review from Anthropic
  • Covers the full document lifecycle from drafting to final file format
  • The skill-creator skill makes it easier to build your own writing skills
  • Works across Claude.ai, Claude Code and the Claude Developer Platform

Cons

  • Not a curated directory, so it won't help you discover community writing skills
  • The writing-specific skills are general-purpose rather than tuned to a particular voice or style

Visit Anthropic Official Skills

#3 VoltAgent awesome-agent-skills

VoltAgent's awesome-agent-skills is a curated GitHub collection of over 1,000 agent skills from official dev teams and the community. According to the repository, it's hand-picked rather than AI-slop generated, and it focuses on skills created and used by actual engineering teams. The collection is compatible with Claude Code, Codex, Antigravity, Gemini CLI, Cursor, GitHub Copilot, OpenCode, Windsurf and more. For writers, the relevant entries include official Claude skills for document handling, internal communications, and brand guidelines, plus community skills for content research, blog writing, SEO, and social media. There are also skills from teams like Corey Haines for marketing and copywriting, and Charlie Hills for social content. The repository is organized by source, so you can see which skills come from official teams and which come from the community. It's a discovery resource, not a verification layer: the README doesn't claim to security-scan every entry, so you should still inspect a skill before installing it. But as a starting point for finding writing-adjacent skills, it's one of the most active and well-maintained collections in the ecosystem.

Best for: Writers and content marketers who want a broad, hand-picked catalog of writing and marketing skills across multiple agent platforms.

Pricing: Free

Key features

  • Curated collection of 1,000+ agent skills from official dev teams and the community
  • Hand-picked rather than bulk-generated, according to the repository
  • Compatible with Claude Code, Codex, Gemini CLI, Cursor, GitHub Copilot, Windsurf and more
  • Includes official Anthropic skills for docx, doc-coauthoring, internal-comms and brand guidelines
  • Community skills for content research, blog writing, SEO and social media from maintainers like Corey Haines and Charlie Hills
  • Organized by source so you can distinguish official team skills from community submissions
  • Actively maintained with contributions from the community

Pros

  • Large, actively maintained collection with a clear distinction between official and community sources
  • Covers writing-adjacent categories like SEO, social content, and marketing copy
  • Works across most major agent platforms that support the SKILL.md format

Cons

  • It's a discovery list, not a security-scanned directory, so you need to review skills yourself before installing
  • Quality varies across community submissions, and the repository doesn't publish benchmark results

Visit VoltAgent awesome-agent-skills

#4 AgenticSkills

AgenticSkills is a curated directory that indexes agent skills for Claude Code, OpenAI Codex, Cursor, Gemini CLI and other SKILL.md-compatible agents. According to the site, it lists 193+ curated skills across 16 categories and tracks 139 MCP servers and 9 platforms. Its featured writing-adjacent picks include Frontend Design by Anthropic, Marketing Skills by Corey Haines (a bundle for CRO, copywriting, SEO, analytics, paid ads and lifecycle email), Claude SEO by Daniel Agrici (an agency-grade SEO bundle with 25 sub-skills and 18 sub-agents), and Social Media Skills by Charlie Hills (a 17-skill social content suite covering hooks, post writing, scoring, reels scripts, carousels and newsletters). Each listing links to the upstream repository, shows the real SKILL.md where one is published, and reports commit and contributor activity scoped to that skill's own path rather than an install count. That commit-level activity signal is useful for judging whether a writing skill is still maintained. The directory also separates skills from MCP servers, which helps you understand whether a listing gives the agent knowledge or a live connection to a tool.

Best for: Writers and marketers who want a smaller, human-curated set of writing and SEO skills with upstream activity signals.

Pricing: Free

Key features

  • Curated directory indexing 193+ skills across 16 categories
  • Tracks 139 MCP servers and 9 agent platforms
  • Featured writing-adjacent skills include Marketing Skills by Corey Haines, Claude SEO by Daniel Agrici, and Social Media Skills by Charlie Hills
  • Each listing links to the upstream repository and shows the real SKILL.md where published
  • Reports commit and contributor activity scoped to the skill's own path, not vanity install counts
  • Separates skills from MCP servers so you know whether a listing is knowledge or a tool connection
  • Covers Claude Code, Codex, Cursor, Gemini CLI and other SKILL.md-compatible agents

Pros

  • Commit-level activity data helps you avoid abandoned writing skills
  • Includes several well-regarded marketing and SEO bundles from known maintainers
  • Clear separation between skills and MCP servers

Cons

  • The catalog is smaller than open directories, so niche writing skills may not be listed
  • It doesn't publish security scan results or benchmark Lift scores for each skill

Visit AgenticSkills

#5 skills.sh

skills.sh is a cross-platform Agent Skills directory with a command-line install flow. According to the AI Agents Library comparison, it's an open ecosystem with a repo-based model, developer-leaning audience, and a public leaderboard with install counts. You can install a skill with a command like npx skills add, which makes it one of the fastest ways to try a writing skill without manually copying files. The catalog is large and community-contributed, and the install counts give you a rough popularity signal. The trade-off is curation: skills.sh is open and minimal on security review, so it's best treated as a discovery tool rather than a trust layer. For writers who are comfortable reading a SKILL.md before installing it, it's a practical way to find and test writing, SEO, and content skills across agents. The CLI install also makes it easy to uninstall or swap skills as you compare them.

Best for: Developers and technical writers who want a fast CLI install flow and a large catalog to browse.

Pricing: Free

Key features

  • Cross-platform Agent Skills directory with CLI install via npx skills add
  • Open ecosystem with a repo-based model
  • Public leaderboard with install counts
  • Large community-contributed catalog
  • Developer-leaning audience and workflow
  • Works across SKILL.md-compatible agents

Pros

  • CLI install makes trying and removing skills quick
  • Install counts give a rough popularity signal
  • Large catalog increases the odds of finding a niche writing skill

Cons

  • Minimal curation and security review, so you inspect before installing
  • Developer-leaning, which means fewer business-writing skills than coding skills

Visit skills.sh

#6 SkillsMP

SkillsMP is an open directory that aggregates SKILL.md files from GitHub. According to the Agentman ecosystem report, it lists about 1.9 million public skills scraped from GitHub, and the AI Agents Library comparison notes that it's aggregated and explicitly not certified. The volume is the point: if a writing skill exists in a public repo, there's a decent chance it's indexed here. That makes SkillsMP useful for discovery when you're looking for something specific, like a skill for a particular style guide or a niche content format. The caveat is that there's no security review. The same ecosystem report cites a security audit of 22,511 skills that found 140,963 issues, and Snyk's ToxicSkills research detected prompt injection in 36% of skills tested. So SkillsMP is a place to find candidates, not a place to trust them. Read the SKILL.md, check the repo's activity, and consider running it through a verification layer before you let it touch your drafts.

Best for: Writers and researchers who need to find a very specific skill and are willing to audit it themselves.

Pricing: Free

Key features

  • Aggregates roughly 1.9 million public SKILL.md files scraped from GitHub
  • Open directory with no certification or security review
  • Useful for finding niche or long-tail writing skills
  • Free to browse and use
  • Covers skills for all professions, though developer-heavy by volume

Pros

  • Enormous catalog makes it likely you'll find an exact match
  • Free and open, with no gatekeeping
  • Useful for discovering skills that smaller directories miss

Cons

  • No security review or certification, so every skill needs manual inspection
  • Catalog size makes quality control impossible, and many entries are unmaintained

Visit SkillsMP

#7 Agensi

Agensi is a curated marketplace for agent skills with an 8-point security scan. According to the Agentman ecosystem report, it has a smaller curated catalog and reviews skills before listing. The Agensi site says every skill goes through an automated checklist covering prompt injection, data exfiltration, secret detection, dangerous commands, obfuscation, external fetches, credential access, and privilege escalation. It also offers creator payments with a 70% revenue share, and both paid and free skills are available. For writers, that means you can find vetted writing, SEO, and content skills with a clearer trust signal than an open scraper provides. The catalog is smaller than SkillsMP or skills.sh, so it won't have everything, but the review process is the differentiator. If you're installing a skill that will read your drafts or access your CMS, an 8-point scan is a reasonable filter before you run it.

Best for: Writers and teams who want a security-scanned marketplace and are willing to pay for vetted skills.

Pricing: Free and paid skills; creator revenue share is 70%

Key features

  • Curated marketplace with an automated 8-point security checklist
  • Checklist covers prompt injection, data exfiltration, secret detection, dangerous commands, obfuscation, external fetches, credential access and privilege escalation
  • Creator payments with a 70% revenue share
  • Both paid and free skills available
  • Works across all SKILL.md-compatible agents
  • Reviewed before listing, according to the site

Pros

  • Published security checklist gives a clear trust signal
  • Creator payments encourage higher-quality submissions
  • Works across multiple agent platforms

Cons

  • Smaller catalog than open directories
  • Paid skills add cost, and free options may be limited

Visit Agensi

#8 ClaudeSkills.info

ClaudeSkills.info is a free directory that aggregates SKILL.md files from GitHub and other sources, with a focus on Claude Code. According to the Agensi directory list, it's one of the largest free directories, with a large catalog but no security scanning or creator verification. The Agentman ecosystem report describes it as community-vetted for Claude Code. That combination means it's useful for discovery, especially if you're looking for writing skills that other Claude users have tried, but you still need to review the code before installing. The community-vetting signal is informal: there's no published scan or benchmark, just other users who have looked at it. For a free directory with a Claude-specific focus, that's a reasonable starting point. Just don't treat a listing as an endorsement. Read the SKILL.md, check what files it bundles, and pay attention to any instructions that tell the agent to connect to external sources.

Best for: Claude Code users who want a free, Claude-focused directory and are willing to review skills manually.

Pricing: Free

Key features

  • Free directory aggregating SKILL.md files from GitHub and other sources
  • Focus on Claude Code and Claude-compatible agents
  • Community-vetted for Claude Code, according to the Agentman ecosystem report
  • Large catalog with no security scanning or creator verification
  • Useful for discovery before manual review

Pros

  • Free and Claude-specific, which narrows the search
  • Community vetting gives a weak but useful signal
  • Large catalog for a niche directory

Cons

  • No security scanning or creator verification
  • Community vetting is informal and not a substitute for reading the skill

Visit ClaudeSkills.info

#9 Smithery

Smithery is the leading directory for MCP servers rather than SKILL.md skills. According to the Agensi directory comparison, it focuses on Model Context Protocol servers, which connect agents to external tools and data sources. That's a different layer from skills: a skill teaches the agent a procedure, while an MCP server gives it a live connection to something like a CMS, a database, or a research API. For writers, Smithery is useful when your writing workflow needs to pull from or push to an external system. A skill might define how to draft a blog post, and an MCP server might give the agent access to your WordPress instance to publish it. The directory lists available servers and their capabilities, so you can find the connection you need. It won't help you find a writing skill, but it's a practical companion to any skills directory when your workflow crosses the boundary from knowledge into action.

Best for: Writers and developers who need to connect their agent to a CMS, database, or research API as part of a writing workflow.

Pricing: Free to browse; individual MCP servers may have their own pricing

Key features

  • Leading directory for Model Context Protocol servers
  • Focuses on connecting agents to external tools and data sources
  • Lists available MCP servers and their capabilities
  • Different layer from SKILL.md skills: connectivity rather than procedure
  • Useful for writing workflows that need to publish or pull data

Pros

  • The most focused directory for MCP servers
  • Complements skills directories by covering the connection layer
  • Useful for automating the publishing step of a writing workflow

Cons

  • Not a skills directory, so it won't help you find writing skills
  • MCP servers can introduce their own security and access considerations

Visit Smithery

Conclusion

Writing with Claude gets a lot better when you stop relying on one giant prompt and start installing skills that encode how you actually work. The nine options above cover the full range: official first-party skills for document handling, curated GitHub collections for discovery, open directories with huge catalogs, a security-scanned marketplace, and an MCP directory for the connection layer. The one thing they don't all cover is trust. Research has found vulnerabilities in a significant share of published skills, and a prompt-injection payload hidden in a writing skill can do real damage. That's why SkillzBench sits at number one: it's free, it pins every skill to a commit SHA, it scans and benchmarks before listing, and it publishes a Trust score and a Lift result so you can see whether a writing skill actually improves output. Browse hundreds of writing and editing skills at skillzbench.com, connect your agent through the free MCP server, and toggle only the skills you want it to see. Skills, proven.

Frequently asked questions

What is a Claude Skill, and how is it different from a prompt?

A Claude Skill is a folder containing a SKILL.md file with metadata and instructions, and optionally scripts, references, and assets. According to Anthropic's engineering blog, the agent pre-loads only the name and description of each installed skill, then reads the full instructions when a task matches. A prompt is a one-off instruction you type into a conversation. A skill is a persistent, version-controlled artifact that any compatible agent can load on demand. That difference matters for writing because it lets you keep a whole library of style guides, formats, and workflows available without stuffing every conversation with thousands of tokens of instructions.

Are Claude Skills safe to install?

Not automatically. Skills can include instructions and executable code, and Anthropic's own guidance recommends installing only from trusted sources and auditing anything from a less-trusted one. The risk is real: Snyk's ToxicSkills research detected prompt injection in 36% of skills tested, and a security audit of 22,511 skills found 140,963 issues. A malicious writing skill could instruct the agent to exfiltrate your drafts or connect to an untrusted external source. The practical answer is to use a verification layer. SkillzBench pins every skill to a commit SHA, runs security and license checks, and publishes a Trust score so you can see what's been reviewed before you install.

How does SkillzBench benchmark a writing skill?

SkillzBench tests every skill in a sandbox with the same model, running the task with and without the skill installed, and publishes the difference as a Lift percentage. For a writing skill, that means you can see whether it actually improves the output rather than just changing it. The Trust score is shown separately from Lift, and money never touches either. Everything on the platform is free, including the benchmarks, and the company earns revenue from advertising around its blog articles, not from charging users or influencing scores.

Do I need to pay for any of these skill directories?

Most of them are free. SkillzBench, Anthropic Official Skills, VoltAgent awesome-agent-skills, AgenticSkills, skills.sh, SkillsMP, ClaudeSkills.info and Smithery are free to browse, and SkillzBench's directory, benchmarks and MCP connector are free as well. Agensi offers both free and paid skills, with a 70% revenue share for creators. Some individual MCP servers listed on Smithery may have their own pricing. If you're using Claude itself, you'll still need a Claude plan to run the skills.

Can I use these skills with agents other than Claude?

Many of them, yes. The Agent Skills format was published as an open standard, and according to the Agentman ecosystem report, roughly 40 skills-compatible products appear on the official showcase, including OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI and VS Code. VoltAgent's collection explicitly lists compatibility with Claude Code, Codex, Antigravity, Gemini CLI, Cursor, GitHub Copilot, OpenCode and Windsurf. SkillzBench's MCP connector is designed to work with any agent, and its per-agent loadouts let you choose which skills each agent sees.

What's the difference between a skill and an MCP server?

A skill gives the agent knowledge: how to perform a task, what format to use, what rules to follow. An MCP server gives the agent access: a live connection to a database, a CMS, or an API. As the Goose team at Block put it, skills describe the workflow while MCP provides the runner. For writing, a skill might define how to draft a blog post in your voice, and an MCP server might let the agent publish it to your CMS. Most real setups use both. Smithery is a directory for the MCP side, and SkillzBench provides a free MCP connector that links your agent to the skills you select.

How do I know if a writing skill is still maintained?

Look for commit and contributor activity scoped to the skill's own path, not just the repository's overall star count. AgenticSkills reports that kind of activity for each listing, and SkillzBench tracks upstream changes and re-scans and re-benchmarks skills when they change, then notifies you. A high star count on a monorepo doesn't tell you whether the writing skill you want has been touched in a year. A recent commit on the skill's own directory does. If a directory doesn't show that signal, check the upstream repo before you install.