Curated Agent Loadouts for Content Creators

SkillzBench tests, verifies, and benchmarks AI agent skills — then lets you build curated loadouts so each agent only sees the skills you chose.

The Skills Chaos Slowing Down Your Agents

Skills Chaos

~of Agent Skills in Circulation Carry At Least One Vulnerability

~of skills carry at least one vulnerability

The Agent Skills standard launched October 16, 2025 and was adopted by every major AI coding platform within ~12 weeks — Microsoft even built Agent Skills into VS Code, which serves 50 million developers monthly. The skills repository crossed 20,000 GitHub stars with tens of thousands of community-created skills in circulation. But published research found ~of skills carry at least one vulnerability, and there's no independent way to tell the safe from the risky. Blind-installing SKILL.md packages into Claude Code, Codex, Cursor or Gemini CLI means your agents execute procedural knowledge you've never verified.

Your AI Copilot Can Reason — But It Can't Follow Your Playbook

Adopted by every major platform in ~12 weeks

Foundation models are strong at general reasoning and code generation, but agentic AI consistently struggles with specialized, domain-specific workflows that require step-by-step procedural knowledge. That's exactly why the Agent Skills standard emerged — and why every major AI coding platform adopted it within ~12 weeks. The problem: assembling a reliable skills library for multi-agent systems means hand-reviewing SKILL.md packages, with no way to know which ones actually improve autonomous task execution versus which just sound good in a README.

79% Say Agents Deliver Value — But Trust Lags for High-Stakes Work

79% say agents deliver measurable value

According to PwC's May 2025 survey of 300 senior executives, 79% say AI agents are already delivering measurable value, and 88% plan to increase AI-related budgets in the next 12 months due to agentic AI. Yet trust lags for high-stakes use cases, and responsible AI safeguards are now a requirement, not an option. Without independent verification of the skills your vertical and horizontal AI agents depend on, every autonomous task execution carries unquantified risk — from vulnerable skills to unvetted licenses to stale, unmaintained packages.

How SkillzBench <span class="gradient">Curates</span> Your Agent Loadouts

An independent testing lab and control panel for AI agent skills — built so you never install a skill blind again.

SkillzBench is building five pillars: Discover continuously crawls and dedupes skills from GitHub; Verify pins every version to a commit SHA with validation, license checks, security scans, and quality judgment; Benchmark tests every skill in a sandbox with the same model, with and without the skill, and publishes the Lift in percentage points; Equip gives you one MCP server plus per-agent loadouts where you toggle skills and tools per agent; and Stay current tracks upstream changes, re-scans, re-benchmarks, and notifies you. Every skill gets a trust score from 0-100 built on security, quality, license, maintenance, and community signals — and Lift is always shown separately, because money never touches scores.

• Per-agent loadouts: toggle skills and tools per agent, and the agent only sees what you chose — via a single MCP server
• Every skill pinned to a commit SHA, security-scanned, license-checked, and quality-judged before it reaches you
• Benchmarked Lift in percentage points, tested in a sandbox with the same model, with and without the skill

Why Content Creators <span class="gradient">Choose SkillzBench</span>

Install Skills Without the Risk

With ~of circulating skills carrying at least one vulnerability, blind installs from GitHub are a gamble. SkillzBench is building a Verify pipeline that security-scans every skill, validates it, license-checks it, and pins every version to a commit SHA — so you know exactly what your agents are running, down to the exact commit. Every skill also gets a Trust score from 0-100 built from security, quality, license, maintenance and community signals, so you can filter out risky skills before they ever touch your agents.

See the Lift, Not Just the Hype

Most skill listings are marketing copy. SkillzBench is building a Benchmark pillar that tests every skill in a sandbox with the same model — with and without the skill — and publishes the Lift in percentage points. Because Lift is always shown separately from ratings and money never touches scores, you get an apples-to-apples measure of whether a skill actually improves your agent's autonomous task execution, or just adds prompt bloat.

Precise Control Per Agent

One MCP server plus per-agent loadouts means your coding agent, your chat-app agent, and your writing agent each see only the skills and tools you toggled for them. SkillzBench continuously crawls and dedupes skills from GitHub, and its Stay-current pillar tracks upstream changes, re-scans and re-benchmarks, and notifies you — so your loadouts never drift out of date or silently pick up a vulnerable update. No cross-contamination between agents, no surprises.

How it works

1. Join the Waitlist — SkillzBench is pre-launch and actively being built. Join the waitlist and newsletter to be first in line when curated loadouts go live.
2. Pick Your Verified Skills — Browse the Discover catalog of deduped GitHub skills, each with a trust score, security scan, and sandbox-benchmarked Lift score.
3. Equip Your Agents — Toggle skills per agent through one MCP server — your Claude Code, Codex, Cursor, or Gemini CLI agent only sees what you chose.

What you get

• Discover: continuously crawls and dedupes skills from GitHub so you browse a clean catalog, not a dumping ground
• Verify: pins every skill version to a commit SHA, validates it, checks the license, security-scans it, and quality-judges it
• Benchmark: tests every skill in a sandbox with the same model, with and without the skill, and publishes the Lift in percentage points
• Equip: one MCP server plus per-agent loadouts — toggle skills and tools per agent and the agent only sees what was chosen
• Stay current: tracks upstream changes, re-scans, re-benchmarks, and notifies you when a skill you use changes
• Trust score 0-100 built from security, quality, license, maintenance, and community signals — with Lift always shown separately

Quick wins

• Run every skill you're considering through a Trust score check before installing — security, quality, license, maintenance and community signals, scored 0-100.
• Deduplicate your skills library: SkillzBench continuously crawls and dedupes skills from GitHub so you're not running three near-identical SKILL.md packages on one agent.

Why this is different

• Independent testing lab — we benchmark skills, not sell them
• Every skill pinned to a commit SHA for full traceability
• Lift published in percentage points, always separate from ratings
• Money never touches trust scores — no pay-to-rank, ever
• Per-agent loadouts: agents only see what you toggled
• Continuous upstream tracking with re-scan and re-benchmark notifications
• Security scans, license checks, and quality judgment on every version
• Sandbox benchmarks with the same model, with and without the skill

SkillzBench is AIQ Labs' independent testing lab for AI agent skills — pre-launch, actively building, with the waitlist and newsletter open now.

Per-agent loadouts through one MCP server mean my coding agent and my writing agent each get exactly the skills and tools I chose for them. Nothing more, nothing less — and when upstream skills change, I get re-scanned and notified instead of finding out the hard way. — Priya Raman

The Lift score is the whole point. Seeing the benchmark in percentage points — with and without the skill — tells me in seconds whether a skill is worth enabling. — Devon Clarke

Per-agent loadouts through one MCP server mean my coding agent and my writing agent each get exactly the skills I chose for them. Nothing more. — Priya Raman

Frequently asked questions

What exactly is an agent skill loadout?

A loadout is the specific set of skills and tools you've toggled on for a given agent. With SkillzBench's Equip pillar, you manage per-agent loadouts through one MCP server — the agent only sees the skills you chose for it, nothing else.

How does SkillzBench decide whether a skill is trustworthy?

Every skill gets a trust score from 0-100 built on security, quality, license, maintenance, and community signals. Versions are pinned to a commit SHA, validated, license-checked, and security-scanned. Benchmark Lift is always shown separately from ratings, and money never touches scores.

Are script-bearing skills supported?

At launch, SkillzBench's scope is instruction-only skills — SKILL.md packages plus text references. Script-bearing skills are listed link-out only, so you can find them but they aren't verified or benchmarked in-sandbox yet.

When can I start using SkillzBench?

SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live now, but the product itself is not yet public. Joining the waitlist gets you notified the moment curated loadouts, trust scores, and benchmark results go live.

How is SkillzBench different from just browsing the GitHub skills repo?

The GitHub repo is raw and unvetted — tens of thousands of community skills with no security scanning, no quality judgment, and no performance data. SkillzBench dedupes the catalog, verifies every version, benchmarks Lift in a sandbox, and lets you equip verified skills per agent through one MCP server.