Jonathon's AI Wiki

Claude Skills Hub (claudeskills.info)

8 min read

Source: Claudeskills Info Directory 2026 04 26

URL: https://claudeskills.info/ Explore: https://claudeskills.info/explore/ Total catalog (claimed): 658+ skills (at time of ingest 2026-04-26) Featured on: Product Hunt

A third-party community marketplace and search directory for Claude Skills. Aggregates skills from official publishers (Anthropic, OpenAI, Microsoft, Google, Vercel, GitHub, WordPress) alongside community-built collections like Trail of Bits security research, MITRE-mapped cybersecurity skills, Jesse Vincent’s Superpowers, and curator bundles. Distinct from the wiki’s existing references to SkillsMP, TokRepo, and awesome-claude-skills because it foregrounds cross-vendor official collections — making it the easiest single place to compare ecosystem maturity across LLM vendors that implement the Agent Skills open standard.

Key Takeaways

  • Cross-vendor official collections are the differentiator. Anthropic (16), OpenAI (37), Microsoft (333), Google (11), Vercel (8), GitHub Copilot (324), WordPress all surface as first-class publisher collections. No other Claude-skills directory the wiki currently tracks does this.
  • 658+ skills total. Substantially smaller than SkillsMP’s 800k+ but with a higher signal-to-noise ratio — most listings are publisher-vetted or curator-vetted rather than raw user uploads.
  • 12 categories with skill counts. Development (83), Productivity (60), AI & ML (23), Office & Documents (13), Creative & Art (12), Data & Analysis (11), Design & Styling (9), Testing (9), Communication (8), Security (5), Meta & Tools (3), plus Learning. Useful for scoped browsing.
  • Surfaces well-known community packs. Cybersecurity Skills (734+, MITRE-mapped, mukul975), Trail of Bits Security Research (45 skills), Everything Claude Code (86+, affaan-m, 76.2k), PM Skills Marketplace (63, phuryn). Each links out to the full bundle.
  • Cross-references match the wiki’s existing graph. Anthropic’s 16 official skills correspond to skills (the wiki tracks the repo at 17, the Hub shows 16 — minor counting drift, likely template-skill exclusion). Superpowers shows up here at 1.2k via the Hub vs the repo’s actual 168k stars — the Hub’s counter is its own engagement metric, not GitHub stars.
  • Fetched via Tavily. WebFetch was blocked (HTTP 403, CDN bot challenge). Snapshot was extracted via Tavily MCP fallback — same operational pattern documented for ingest routines.
  • Install/submission mechanism not visible from the public site snapshot. The Hub may have an internal install flow or PR-based submission, but the explore page doesn’t surface it. Treat as a discovery surface; install through the underlying GitHub repo or Claude Code marketplace.
  • Trust gradient runs publisher → curator → individual. Use Hub → Anthropic / OpenAI / Microsoft collections for production-grade skills, curator bundles (Trail of Bits, mukul975) for specialty domains, individual skills with caution.

Official publishers

PublisherSkills in collectionDomains covered
Anthropic16Frontend design, PDF, DOCX, XLSX, MCP builder, document tools
OpenAI37Codex CLI, multi-agent development
Microsoft333Azure, .NET, AI Foundry — largest enterprise collection
Google11Google Labs, Gemini, Stitch design
Vercel8Next.js, React, edge deployment
GitHub Copilot324Azure, .NET, Docker
WordPress(count not surfaced)WordPress plugin skills

Community / partner highlights

  • Cybersecurity Skills (mukul975) — 734+ skills mapped to MITRE ATT&CK; the largest single security-domain bundle
  • Trail of Bits Security Research — 45 skills focused on fuzzing, static analysis, vulnerability detection
  • Everything Claude Code (affaan-m) — 86+ skills, rules, hooks, agent configs; 76.2k engagement counter
  • PM Skills Marketplace (phuryn) — 63 product-management skills covering discovery → strategy → execution → launch → growth
  • Superpowers (Jesse Vincent) — listed at 1.2k on the Hub; corresponds to Superpowers (168k stars on GitHub)

How it compares to the wiki’s other directory references

SourceScaleVettingWhat it’s good for
claudeskills.info (this article)658+ skills, 12 categories, multi-vendor official collectionsMixed: official publisher tier + curator tier + communityComparing official publisher coverage; finding curator bundles
skills17 skillsAnthropic-published referenceAuthoring skills; reading the spec; trusted starter pack
SkillsMP (per plugins-and-marketplaces)800,000+ entriesCommunity uploads, no formal vettingMaximum-coverage discovery; long-tail searches
TokRepo (per plugins article)500+ skills, MCP servers, workflowsCurated for discoveryCross-primitive search (skills + MCP + workflows together)
awesome-claude-skills (per plugins article)1,234+ skillsCommunity ratings and reviewsRead-the-reviews discovery

The Hub’s strength is the publisher collections layer — none of the others foreground cross-vendor official catalogs.

Sample skills surfaced in /explore/

(Top of stream at time of ingest; the catalog updates continuously.)

  • OpenClaw Chrome Relay Helper (i-am-rad, Frontend) — Pre-req for browser automation requiring OpenClaw Browser Relay Chrome extension.
  • Polymarket Screener (bytesagain1) — Screen prediction markets by category, probability, liquidity.
  • Cost Control 3-Tier API Spend Protection (theshadowrose, AI/ML) — Caution / emergency / hard-cap protection across multi-provider API spend.
  • Token Tamer — AI API Cost Control (theshadowrose) — Multi-provider monitor with budget enforcement.
  • CurlShip Directory Submission (marcindudekdev) — Single-curl SaaS-directory submission with OG-tag scraping.
  • Transcribe (dairui1) — Podcast → audio + SRT subtitles + TXT transcript.
  • Trugen AI — Build, configure, deploy conversational video agents.
  • Vercel React Best Practices (Frontend) — React/Next.js performance optimization.
  • Semgrep Rule Variant Creator (Testing) — Port Semgrep rules to target languages with rule + test pairs.
  • Word Document Processor / PowerPoint Processor / Excel Spreadsheet Processor (Anthropic, Office & Docs, 5.3k each) — The production document skills surfaced as discrete cards.

Each card carries:

  • One- to two-sentence description with usage trigger (“Use when …” / trigger phrases)
  • Author handle / publisher
  • Category badge + free-form tag chips
  • Engagement counter (views? installs? unclear from snapshot)

When to use it

  • You want to compare official skill coverage across vendors (Anthropic vs OpenAI vs Microsoft vs Google) at a glance.
  • You’re researching what specialty bundles exist for a domain — security, PM, frontend, etc. — without trawling individual GitHub repos.
  • You want to discover community skill-bundle creators whose work might be worth following.
  • You’re building a curated skills shortlist for a team and need a second opinion on which collections are visible.

When not to use it

  • You need an authoritative install path — go through anthropics/skills (plugin marketplace) or the underlying GitHub repo. The Hub is for discovery, not necessarily install.
  • You need vetted production skills only — pre-filter to Anthropic / OpenAI / Microsoft official tiers; treat individual community skills as read-and-test before relying on them.
  • You’re writing skills for a single workflow — template/ in skills plus the Building Skills Guide is a faster path than the Hub’s discovery flow.

Tradeoffs

  • Counts can drift. The Hub’s per-collection counts and the underlying GitHub repos’ counts don’t always match (16 vs 17 for Anthropic, 1.2k vs 168k for Superpowers). Trust the GitHub source for stars; trust the Hub for relative scale comparisons.
  • No visible quality vetting. The site doesn’t publish a moderation policy. Publisher tiers are presumably trusted by virtue of the publisher; community tier is on you.
  • Install mechanism unclear from public surface. Discovery is good; the install handoff isn’t documented from the explore page. Plan to bridge to GitHub or /plugin install.
  • CDN-blocked from automation. The Hub returns HTTP 403 to direct WebFetch, requiring Tavily fallback — fine for ad-hoc lookups, friction for any pipeline that wants to scrape it.

Try It

  1. Open https://claudeskills.info/explore/ and scan the publisher-collection cards. Note which vendors have how many skills — useful baseline for ecosystem maturity.
  2. Filter to one category that maps to current work (e.g., Security if doing pentest research, Office & Documents if shipping reports). Read 3-5 skill cards to calibrate quality.
  3. For any skill that looks promising, click through to its source repo before installing. Verify license, last-update date, and SKILL.md quality firsthand.
  4. Cross-reference any community pack you’re considering against the wiki’s existing entity articles (Superpowers, Marketing Skills Bundle, etc.) — if it’s already documented, lean on the wiki article. If not, propose a Research operation.
  5. Don’t rely on the Hub’s install flow — install via /plugin install or npx skills add per the underlying repo’s documented path.

Open Questions

  • What’s the moderation policy? Are there reported takedowns of low-quality or malicious skills?
  • Is there an API or RSS feed for new skill listings? (Useful for adding to the wiki’s watchlist routine if so.)
  • How is the Hub funded — affiliate, advertising, sponsored placements? The “third-party marketplace” framing leaves this unspecified.
  • What’s the submission flow? Is it PR-based (like awesome lists) or form-based (like a SaaS marketplace)?
  • The 658+ total vs ~250 directly-categorized count gap — what fills the difference? Probably collection cards being counted at the publisher level rather than per-skill, but worth verifying.

Graph View

Backlinks