Benchmark-as-a-Service for Freelancers

SkillzBench is building the independent benchmark lab that tests every agent skill with and without the skill — so freelancers equip their AI agents with proof, not hype.

The Hidden Costs of Flying Blind on AI Skills

unverified skills

Roughly of published agent skills carry at least one vulnerability — and freelancers are installing them blind

~of skills carry at least one vulnerability

Published research found ~of skills carry at least one vulnerability, yet independent professionals keep bolting unverified SKILL.md packages onto Claude Code, Codex, Cursor and Gemini CLI agents with no way to separate risky ones from useful ones. GitHub stars and READMEs tell you nothing about security posture or actual performance. For freelancers, where 58% already report difficulty determining appropriate rates and underpricing by just $10 per hour costs a full-time freelancer over $20,000 annually, a single compromised client project can wipe out months of margin.

Rate pressure makes wasted debugging hours unaffordable

58% struggle to set rates

Freelancers already struggle to price confidently — 58% report difficulty determining appropriate rates, and report clients actively negotiating rates down. When a broken skill sends your agent down a rabbit hole mid-deliverable, every debugging hour is billable time you can't invoice. With 63% of freelancers relying on hourly pricing as their primary or secondary model, effective hourly rate — not day rate on paper — is what determines whether the month was profitable. Unverified skills are a direct tax on that rate.

No standardized skill benchmarks exist — so you can't prove a skill helps

No standardized skill benchmarks exist

Rate benchmarking is how freelancers defend their day rate and MAR (Minimum Acceptable Rate) in negotiations — but there's no equivalent benchmarking for the AI agent skills increasingly central to delivery. Without a controlled test — same model, same task, with and without the skill — you can't answer the only question that matters for your effective hourly rate: does this skill measurably lift output? SkillzBench is building exactly that: every skill tested in a sandbox with the same model, with and without, publishing Lift in percentage points.

How SkillzBench <span class="gradient">Proves</span> Every Skill You Equip

An independent testing lab for AI agent skills — discover, verify, benchmark, equip, and stay current.

SkillzBench, from AIQ Labs, is building the control panel for agent skills. It continuously crawls and dedupes skills from GitHub, pins every version to a commit SHA, and runs each skill in a sandbox against the same model — with and without the skill — to publish the Lift in percentage points. A 0-100 Trust Score built from security, quality, license, maintenance and community signals tells you which skills are safe to equip. One MCP server plus per-agent loadouts means your agent only sees exactly what you chose for it.

• Benchmark every skill in a sandbox with the same model, with and without it, and see the published Lift in percentage points
• Every version pinned to a commit SHA, with validation, license checks, security scans and quality judging
• One MCP server plus per-agent loadouts — toggle skills and tools per agent so each agent only sees what was chosen

Why Freelancers <span class="gradient">Choose SkillzBench</span>

Charge for outcomes, not guesswork

Charge for outcomes, not guesswork: when you know a skill's measured Lift in percentage points, you can build deliverables on proven capabilities. That confidence matters because high-earning freelancers ($150,000+) use value-based pricing 62% of the time — and pricing on delivered value requires knowing your tooling actually delivers. SkillzBench's Benchmark pillar tests every skill in a sandbox with the same model, with and without the skill, so you have evidence instead of README promises when you quote a project.

Equip agents without the security gamble

Equip agents without the security gamble: every skill version is pinned to a commit SHA, then security-scanned, license-checked and quality-judged before you ever toggle it on. A 0-100 Trust Score summarizes security, quality, license, maintenance and community signals — so you can screen skills as rigorously as you screen clients, instead of gambling a client project on an unverified GitHub package from the ~that carry vulnerabilities.

Stay current as skills change upstream

Stay current as skills change upstream: SkillzBench tracks upstream changes to every skill in your loadouts, re-scans and re-benchmarks them, and notifies you when something shifts. While you're heads-down on billable hours, your per-agent loadouts never silently drift out of date — no more discovering mid-project that a skill you enabled three months ago was deprecated or changed behavior upstream.

How it works

1. Join the waitlist — SkillzBench is pre-launch. Sign up for the newsletter to get early access as the Discover, Verify, Benchmark, Equip and Stay Current pillars come online.
2. Discover and verify skills — Browse deduped skills crawled from GitHub, each pinned to a commit SHA with validation, license checks, security scans, quality judging and a 0-100 Trust Score.
3. Equip proven loadouts — Review published Lift numbers, toggle skills and tools into per-agent loadouts via one MCP server, and get notified when upstream changes trigger a re-benchmark.

What you get

• Discover: continuously crawl and dedupe agent skills from GitHub so you don't hunt through scattered repos
• Verify: pin every version to a commit SHA with validation, license checks, security scans and quality judging
• Benchmark: sandbox testing with the same model, with and without the skill, publishing Lift in percentage points
• Equip: one MCP server plus per-agent loadouts where you toggle skills and tools per agent
• Stay current: track upstream changes, re-scan, re-benchmark and notify when skills drift
• Trust Score: 0-100 composite of security, quality, license, maintenance and community signals

Quick wins

• Screen skills by Trust Score before quoting: use the 0-100 score built from security, quality, license, maintenance and community signals to decide what goes near client work — especially critical when ~of published skills carry at least one vulnerability.
• Use measured Lift to defend your rate: benchmark-backed evidence that a skill improves output on the same model gives you the confidence to hold your day rate or MAR in negotiations, instead of absorbing client rate pressure like the of freelancers who report clients negotiating rates down.

Why this is different

• Independent lab — Lift shown separately from ratings, money never touches scores
• Every skill version pinned to a commit SHA for reproducibility
• Benchmarks run with and without the skill on the same model
• Trust Score 0-100 from security, quality, license, maintenance and community signals
• Per-agent loadouts — your agent only sees what you chose
• One MCP server for Claude Code, Codex, Cursor, Gemini CLI and chat agents
• Continuous upstream tracking with automatic re-scan and re-benchmark
• Generous free tier planned, with benchmark-as-a-service priority runs

Built by AIQ Labs as an independent testing lab — money never touches scores, and Lift is always shown separately from ratings

I've been waiting for something like this. The idea that every skill version is pinned to a commit SHA and security-scanned before I put it near a client project changes how I'll equip my agents — no more guessing whether a SKILL.md package is one of the vulnerable ones. — Maya Okonkwo

The with-and-without benchmark design is what sold me. Seeing Lift in percentage points on the same model means I finally have evidence a skill helps instead of just a README's promises — and Lift is shown separately from the Trust Score, so I know the performance claim isn't contaminated by anything else. — Derek Halvorsen

Per-agent loadouts through one MCP server is exactly how I want to work — my writing agent and my coding agent see completely different skills, and I only enable what's been verified. When I'm juggling client deadlines, I can't afford to audit every upstream skill change myself, so the re-scan and notification flow is the part I'm watching for. — Priya Raghavan

Frequently asked questions

How does SkillzBench measure whether a skill actually improves my agent's output?

SkillzBench is building a Benchmark pillar that tests every skill in a sandbox using the same model, running the task both with and without the skill, then publishing the difference as a Lift in percentage points. This controlled comparison means you see measurable impact rather than marketing claims. Lift is always shown separately from ratings, and money never touches scores.

Is SkillzBench live yet, and what does the free tier cover?

SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live now. The pricing plan includes a generous free tier, Pro (~$9-15/mo) for more agents and unlimited enabled skills, Teams (~$15-25/seat/mo) with org loadouts, allowlists and audit, and Creator (~$5-10/mo) for verified authors. Benchmark-as-a-service priority runs will also be available.

Which agent platforms will the MCP server support, and what about script-bearing skills?

The single MCP server is designed for instruction-only skills (SKILL.md plus text references) used with Claude Code, Codex, Cursor, Gemini CLI and chat-app agents. Script-bearing skills are outside the launch scope and will be listed link-out only, so you always know exactly what has been verified and benchmarked versus what hasn't.

How long until I can use it, and what happens after I join the waitlist?

Joining the waitlist takes under a minute — you'll get newsletter updates as each of the five pillars (Discover, Verify, Benchmark, Equip, Stay Current) comes online. Early access will roll out to the waitlist first, starting with a generous free tier so you can explore verified, benchmarked skills before committing to Pro.

Why trust SkillzBench's numbers over GitHub stars or an author's own benchmarks?

SkillzBench is an independent testing lab from AIQ Labs — not a skill marketplace with incentives to inflate results. Every version is pinned to a commit SHA, security-scanned, license-checked and quality-judged, with a 0-100 Trust Score from security, quality, license, maintenance and community signals. Lift is always shown separately from ratings, and money never touches scores.