Agenticskills.io vs SkillzBench: Which Agent Skills Platform Should You Trust?
Agenticskills.io is a free, editorially curated directory of 193+ agent skills and 196 MCP servers across 9 platforms, backed by an in-depth learning library — ideal if you want discovery and guides today. SkillzBench (pre-launch, waitlist open) is also free and adds what a directory alone can't: version pinning to commit SHAs, security scanning, sandbox benchmarking with published Lift scores, and a free MCP connector with per-agent loadouts. If trust and verification are the deciding factor, SkillzBench's approach is built for it; if you want to browse skills and read guides right now, Agenticskills.io is live.
TL;DR
- Agenticskills.io indexes 193+ curated agent skills and 196 MCP servers across 9 platforms, with 100% open-source listings (as of October 13, 2026).
- SkillzBench is free end to end — directory, MCP connector, benchmarks and alerts — and earns revenue only from blog advertising that never touches scores.
- Published research found roughly 26% of agent skills carry at least one vulnerability — the core problem SkillzBench's verification pipeline targets.
- Agenticskills.io's own audit coverage found 73% of public MCP servers fail basic security checks across 8 independent audits of 23,000+ servers.
- Agenticskills.io listings are editorial judgment, not measurements — the site states listing is never paid for.
- SkillzBench pins every skill version to a commit SHA and benchmarks each skill in a sandbox with and without it, publishing Lift in percentage points.
- Agenticskills.io is live today; SkillzBench is pre-launch with a live waitlist and newsletter.
Head to head
| Dimension | SkillzBench | Agenticskills.io |
|---|---|---|
| Pricing | Free — directory, MCP connector, benchmarks, alerts; ad-supported blog only | Free newsletter; listings never paid for |
| Catalog & coverage | Continuously crawls and dedupes GitHub skills; pre-launch | 193+ skills, 196 MCP servers, 9 platforms, 16 categories (Oct 2026) 🏆 |
| Trust & verification | SHA-pinned versions, security scans, license checks, quality judgment, 0–100 trust score 🏆 | Editorial judgment and stated methodology; no per-skill scans |
| Benchmarks | Sandbox tests with and without each skill, publishing Lift in percentage points 🏆 | None — ranks are editorial, not measurements |
| Agent integration | Free MCP server plus per-agent loadouts 🏆 | None — browse and install yourself |
| Learning resources | Blog for adopters, builders and prompt engineers | 177+ minutes of guides, 41K+ words, CLI docs verified against skills 1.7.0 🏆 |
| Availability | Pre-launch; waitlist and newsletter live | Live, indexed as of October 13, 2026 🏆 |
| Ongoing monitoring | Tracks upstream changes, re-scans, re-benchmarks, notifies 🏆 | Newsletter reports additions and rejections |
The contenders
Agenticskills.io is a curated directory and learning hub for the AI agent skills ecosystem. As of October 13, 2026 it indexes 193+ curated agent skills across 16 categories and 196 MCP servers, tracking 9 platforms including Claude Code, OpenAI Codex, Cursor and Gemini CLI, with 100% of listed skills open source. Every listing carries an S/A-style rank, install counts, platform compatibility tags and an 'at a glance' block covering vendor, license, deployment and pricing model — with the site explicitly noting that listings reflect editorial judgment, not measurements, and are never paid for. Beyond discovery, the site runs a substantial learning library: 177+ minutes of guides totaling 41K+ words, covering the npx skills CLI (verified against skills 1.7.0), Claude Skills fundamentals, the Agent Plugins open standard backed by OpenAI, Microsoft, AWS, GitHub, Cursor and Vercel, and security analysis of MCP servers. A free newsletter reports what was added to the directory, what was rejected and why.
pricing 0
- + Live and fully browsable today, with a large indexed catalog (193+ skills, 196 MCP servers)
- + Strong editorial layer: ranks, install counts, categories and platform compatibility make comparison easy
- + Deep, current guides — including CLI documentation verified against skills 1.7.0
- + Transparent methodology: listings are stated to be editorial judgment and never paid for
- + Covers security education, including a detailed MCP server audit framework
- − Listings are editorial judgment, not measurements — no per-skill security scans or sandbox benchmarks
- − No tooling to connect skills to your agents; you install and manage skills yourself
- − No continuous re-scanning or change alerts on individual skills as upstream repos change
- − Ranks and picks don't separate measured performance (Lift) from editorial opinion
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. It exists because agent skills — SKILL.md instruction packages for Claude Code, Codex, Cursor, Gemini CLI and chat-app agents — have proliferated faster than anyone can vet them, and published research found roughly 26% of skills carry at least one vulnerability. SkillzBench is built on five pillars: Discover (continuously crawling and deduping skills from GitHub), Verify (pinning every version to a commit SHA, validating, license-checking, security-scanning and quality-judging), Benchmark (testing every skill in a sandbox with the same model, with and without the skill, and publishing the Lift in percentage points), Equip (one free MCP server plus per-agent loadouts, so an agent only sees the skills and tools you toggled on), and Stay current (tracking upstream changes, re-scanning, re-benchmarking and notifying). A 0–100 trust score combines security, quality, license, maintenance and community signals, with Lift always shown separately — money never touches scores. Everything is free; revenue comes from blog advertising only.
pricing 0
- + Verification-first approach: SHA pinning, security scanning and license checks address the ~26% vulnerability rate found in published research
- + Benchmarks measure actual impact — Lift in percentage points, tested with and without each skill under identical conditions
- + Free MCP connector with per-agent loadouts gives fine-grained control over what each agent can access
- + Completely free, with ad revenue structurally separated from scores and listings
- + Continuous monitoring keeps trust scores current as upstream skills change
- − Pre-launch: the product is not yet public, only the waitlist and newsletter are live
- − No published benchmark results or live skill counts to evaluate yet
- − Smaller learning library than an established editorial site while the blog is still growing
Verdict
Winner: none
These two platforms solve different halves of the same problem: Agenticskills.io is a live, curated discovery directory with strong editorial guides, while SkillzBench is building the verification and benchmarking layer that tells you whether a skill is safe and actually helps. Many small businesses will end up using both — browse on Agenticskills.io, verify on SkillzBench.
| Option | Overall | Breakdown |
|---|---|---|
| Agenticskills.io | 7.5 | : 9: 9: 5: 9: 5: 9 |
| SkillzBench | 7 | : 10: 6: 9: 6: 9: 3 |
Trust and verification: the core difference
Pricing compared
Discovery and day-to-day workflow
Which should you pick?
- :
- :
- :
- :
- :
- :
FAQ
Are Agenticskills.io and SkillzBench free to use?
Yes, both. Agenticskills.io offers its directory and newsletter free — the site states the newsletter is 'Free forever' and that listings are never paid for. SkillzBench is free end to end: the directory, the MCP connector, the benchmarks and the change alerts all cost nothing, with revenue coming from advertising around its blog articles that never influences scores or listings.
Why does skill verification matter so much in 2026?
Because the supply of skills has outpaced anyone's ability to vet them. Published research found roughly 26% of agent skills carry at least one vulnerability, and Agenticskills.io's own analysis of 8 independent audits covering 23,000+ public MCP servers found 73% fail basic security checks, with credential leaks, RCE and prompt injection among the failure modes. Installing an unverified skill means trusting an unvetted instruction package inside your agent.
What is SkillzBench's Lift score?
Lift is SkillzBench's benchmark metric: each skill is tested in a sandbox with the same model, once with the skill and once without, and the difference is published in percentage points. Lift is always shown separately from ratings and trust scores, and money never touches scores — so you can see both whether a skill is safe and whether it actually helps.
Can I use both platforms together?
Yes, and for many teams that's the natural workflow. Agenticskills.io is strong on discovery and education — 193+ curated skills, 196 MCP servers and a deep guide library you can use today. SkillzBench adds the verification and delivery layer: SHA-pinned, security-scanned, benchmarked skills connected to your agents through its free MCP server with per-agent loadouts. Browse broadly, then verify before you deploy.
Is SkillzBench available now?
Not yet as a product. SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live, but the platform itself is not yet public. There are no live skill counts or published benchmark results to point to yet; the verification pipeline, sandbox benchmarking and MCP connector are what the team is building toward.
What is an agent skill, exactly?
An agent skill is a folder containing a SKILL.md file that teaches an AI coding agent how to do one thing well. The file carries a name and description in YAML frontmatter, followed by plain-English instructions the agent reads. Skills work across platforms that support the format, including Claude Code, OpenAI Codex, Cursor and Gemini CLI, and the Agent Plugins open standard — agreed by OpenAI, Microsoft, AWS, GitHub, Cursor and Vercel — now standardizes packaging skills and MCP servers together.