Question 1

How does SkillzBench handle script-bearing skills vs instruction-only SKILL.md packages?

Accepted Answer

At launch, SkillzBench's scope is instruction-only skills — SKILL.md files plus text references — which get full verification and benchmarking. Script-bearing skills are listed link-out only, so you get discovery without unverified code execution.

Question 2

What exactly does the published Lift metric measure?

Accepted Answer

Lift is measured by testing every skill in a sandbox with the same model, with and without the skill, and publishing the difference in percentage points. It's always shown separately from ratings and trust scores, and money never touches scores.

Question 3

How do you protect against vulnerable or unmaintained skills getting into my agent's context?

Accepted Answer

Every skill version is pinned to a commit SHA, then validated, license-checked, security-scanned, and quality-judged. Ongoing maintenance signals feed a 0-100 trust score, and upstream changes trigger re-scan, re-benchmark, and notifications.

Question 4

Is SkillzBench live now, and when can I start using the MCP server?

Accepted Answer

SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live, but the product is not yet public. Join the waitlist to be notified when discovery, verification, benchmarking, and the MCP server go live.

Question 5

How is this different from existing MCP directories and skill marketplaces?

Accepted Answer

Directories list thousands of servers and skills with no signal on safety, maintenance, or performance. SkillzBench is an independent testing lab that pins versions, security-scans, benchmarks in a sandbox, and publishes Lift — with money never touching scores, unlike ad- or listing-fee-driven marketplaces.

MCP Server Discovery for Content Creators