Question 1

How does SkillzBench handle skills that contain scripts, not just SKILL.md instructions?

Accepted Answer

At launch, SkillzBench's scope is instruction-only skills — SKILL.md packages plus text references. Script-bearing skills are listed link-out only, so you get discovery and pointers without the platform claiming to have verified or benchmarked code it hasn't.

Question 2

Can we prove to auditors exactly which skill version an agent was running?

Accepted Answer

Yes — every skill version is pinned to a commit SHA as part of the Verify pillar. Combined with per-agent loadouts and Teams-tier audit capability, this gives you a verifiable record of what each agent was equipped with and when.

Question 3

How does the trust score avoid pay-to-play bias like we've seen in other marketplaces?

Accepted Answer

The 0-100 trust score is built from security, quality, license, maintenance and community signals, and money never touches scores. Lift is always shown separately from ratings, so a skill's measured impact is never conflated with its trust rating.

Question 4

What does rollout look like for a security team, and when can we start?

Accepted Answer

SkillzBench is pre-launch and actively being built; the waitlist and newsletter are live now. Once the product ships, onboarding is: connect the MCP server, review verified skills with their trust scores, build per-agent loadouts, and let Stay current handle upstream change tracking from day one.

Question 5

Why use SkillzBench instead of just reviewing skills on GitHub ourselves?

Accepted Answer

Manual review doesn't scale and doesn't persist — SkillzBench continuously crawls and dedupes skills, security-scans and license-checks each version, benchmarks it in a sandbox with the same model, and re-scans and re-benchmarks automatically when upstream changes. Your team gets published Lift and a 0-100 trust score instead of ad-hoc code reviews that go stale.

Team Skill Governance for Writers