Question 1

How does SkillzBench measure whether an agent skill actually improves performance?

Accepted Answer

SkillzBench benchmarks every instruction-only skill in a sandbox using the same model, running tasks both with and without the skill, then publishes the resulting Lift in percentage points. This with/without comparison is the core of the methodology, and Lift is always shown separately from Trust ratings so performance and trust never get blended into one number.

Question 2

Can I trust skills from GitHub, or do I need to worry about vulnerabilities?

Accepted Answer

Published research found roughly of skills carry at least one vulnerability, so caution is warranted. SkillzBench security-scans, license-checks, and quality-judges every skill version, pinning it to a commit SHA, and folds those signals into a 0-100 Trust score alongside maintenance and community signals. Script-bearing skills are listed link-out only at launch, while instruction-only skills get full validation.

Question 3

What happens when a skill I use gets updated upstream?

Accepted Answer

The Stay-current pillar tracks upstream changes, re-scans, and re-benchmarks affected skills, then notifies you. So if a skill's quality or security posture changes after an update, you'll know rather than discovering it through degraded agent output.

Question 4

Is SkillzBench available now, and how do I get early access?

Accepted Answer

SkillzBench is pre-launch and actively being built — the waitlist/newsletter is live, but the product is not yet public. Join the waitlist to be notified as the Discover, Verify, Benchmark, Equip, and Stay-current pillars come online. Planned pricing includes a generous free tier, Pro at roughly $9-15/mo, Teams at roughly $15-25/seat/mo, and a Creator tier at roughly $5-10/mo for verified authors.

Question 5

How is SkillzBench different from a skill marketplace or a leaderboard?

Accepted Answer

SkillzBench is an independent testing lab, not a marketplace — money never touches scores, and Lift is always displayed separately from Trust ratings. Every skill is version-pinned to a commit SHA and tested with and without the skill on the same model, rather than ranked by popularity or paid placement, and per-agent loadouts let you equip only what each agent actually needs.

Agent Skill Benchmarking for Content Creators