Question 1

How does SkillzBench handle MCP server security and tool poisoning risks when I'm evaluating skills for clients?

Accepted Answer

SkillzBench security-scans every skill during verification and folds the results into a 0-100 trust score alongside quality, license, maintenance, and community signals. Every version is pinned to a commit SHA, so you know exactly what was scanned. Note that launch scope covers instruction-only skills (SKILL.md plus text references); script-bearing skills are listed link-out only.

Question 2

What does 'Lift' mean in SkillzBench benchmarks, and why is it separate from the trust score?

Accepted Answer

Lift is the measured difference in performance when the same model runs a benchmark in a sandbox with and without a given skill, expressed in percentage points. It's always shown separately from ratings and trust scores because SkillzBench is an independent lab — money never touches scores, so a score can't be bought.

Question 3

Can I control which skills and tools each of my clients' AI agents can access?

Accepted Answer

Yes — the Equip pillar provides one MCP server plus per-agent loadouts. You toggle skills and tools per agent, and the agent only sees what was chosen. Teams-tier pricing (planned around $15-25/seat/mo) adds org loadouts, allowlists, and audit for multi-client governance.

Question 4

When can I actually use SkillzBench, and what will it cost?

Accepted Answer

SkillzBench is pre-launch and actively being built — the waitlist/newsletter is live now, and the product isn't public yet. The pricing hypothesis is a generous free tier, Pro at roughly $9-15/mo for more agents and unlimited enabled skills, Teams at $15-25/seat/mo for org loadouts and audit, and Creator at $5-10/mo for verified authors, plus benchmark-as-a-service for priority runs.

Question 5

How is SkillzBench different from just browsing the MCP Registry or GitHub for servers?

Accepted Answer

Public registries and GitHub search give you raw listings — research has found more than half of listed MCP projects are invalid or low-value. SkillzBench adds the missing layer: deduped discovery, commit-SHA-pinned verification with license and security checks, sandbox benchmarking with published Lift, per-agent loadouts through one MCP server, and upstream change tracking that re-scans and re-benchmarks automatically.

MCP Server Discovery for Consultants