Question 1

How does SkillzBench verify a SKILL.md skill I've published on GitHub?

Accepted Answer

SkillzBench's Verify pillar pins every version of your skill to a commit SHA, then runs validation, license-checking, security scanning, and quality judging on that exact artifact. Because the version is pinned, the verification results always refer to the precise code you shipped — no ambiguity about what was tested.

Question 2

What does the published Lift score actually measure for my skill?

Accepted Answer

Lift is measured by running the same model in a sandbox twice — once with your skill enabled and once without — and publishing the difference in percentage points. It's always shown separately from ratings and Trust scores, and money never touches scores, so the number reflects performance only, not promotion.

Question 3

Can I submit skills that include scripts, or only instruction-only SKILL.md packages?

Accepted Answer

Launch scope covers instruction-only skills — a SKILL.md file plus text references. Script-bearing skills are listed link-out only, meaning they appear in discovery with a link back to the source rather than being fully verified and benchmarked.

Question 4

How do I get verified as an author, and when can I start?

Accepted Answer

SkillzBench is currently pre-launch and actively being built, so the first step is joining the waitlist. When the product goes live, the Creator tier (~$5-10/mo) will be the path for verified authors to submit skills for verification, security scanning, and benchmarking. Benchmark-as-a-service will also be available for priority runs.

Question 5

How is this different from just posting my skills on GitHub with a README and my own benchmarks?

Accepted Answer

Self-reported benchmarks are exactly what buyers have learned to distrust — reported scores are difficult to verify or compare across tasks, software, and budgets. SkillzBench is an independent lab: it pins versions to commit SHAs, security-scans them, benchmarks with the same model with and without the skill, and publishes Lift separately from a 0-100 Trust score. Money never touches scores, so your numbers carry weight a README never will.

Skill Author Verification for Consultants