Question 1

How does SkillzBench decide which MCP servers and skills make it into the catalog?

Accepted Answer

Discovery continuously crawls and dedupes skills from GitHub, then the Verify pillar filters hard: every version is pinned to a commit SHA, validated, license-checked, security-scanned and quality-judged. Instruction-only skills (SKILL.md plus text references) are in launch scope; script-bearing skills are listed link-out only. Each skill also carries a trust score from 0-100 built on security, quality, license, maintenance and community signals.

Question 2

What does 'Lift' mean and why is it shown separately from ratings?

Accepted Answer

Lift is the measurable improvement a skill provides: SkillzBench benchmarks every skill in a sandbox with the same model, with and without the skill, and publishes the difference in percentage points. It's always shown separately from ratings and trust scores, and money never touches scores — so a sponsored skill can't buy a better number.

Question 3

How does SkillzBench help with MCP tool poisoning and other security vulnerabilities?

Accepted Answer

Tool poisoning is a first-line threat in the MCP ecosystem, and published research found roughly of agent skills carry at least one vulnerability. SkillzBench addresses this at the Verify stage: security scanning, license checks and quality judgment run on every version before it reaches your shortlist, and every version is pinned to a commit SHA so what you reviewed is what you install.

Question 4

When can I actually start using SkillzBench?

Accepted Answer

SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live now, and the product itself is not yet public. Joining the waitlist puts you first in line as the five pillars (Discover, Verify, Benchmark, Equip and Stay current) roll out, and we don't claim live user counts or published benchmark results until they're real.

Question 5

Why not just use existing MCP directories like MCP.so or Glama.ai?

Accepted Answer

Directories catalog — they don't verify or benchmark. A 14-day measurement study found more than half of listed MCP projects are invalid or low-value, and directories don't tell you whether a server actually improves your agent's output. SkillzBench is an independent testing lab: it dedupes the catalog, security-scans and pins versions, publishes Lift in percentage points, and gives you per-agent loadouts on one MCP server.

MCP Server Discovery for Freelancers