Question 1

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

Accepted Answer

SkillzBench benchmarks every skill in a sandbox using the same model, running tasks with and without the skill, and publishes the difference as a Lift in percentage points. Lift is always shown separately from ratings and trust scores, so you see raw performance impact independent of quality or security judgments.

Question 2

What does SkillzBench's security scanning cover for SKILL.md packages?

Accepted Answer

The Verify pillar pins every skill version to a commit SHA, validates its structure, checks its license, runs security scans and applies a quality judgment. This matters because published research found roughly of skills carry at least one vulnerability, so unverified community skills are a real supply-chain risk for startup teams.

Question 3

Can SkillzBench manage skill loadouts across different agent harnesses like Claude Code, Cursor and Gemini CLI?

Accepted Answer

Yes — the Equip pillar provides one MCP server plus per-agent loadouts. Your team toggles which skills and tools each agent gets, and the agent only sees what was chosen. Launch scope covers instruction-only skills (SKILL.md plus text references); script-bearing skills are listed link-out only.

Question 4

When can my team start using SkillzBench, and what will pricing look like?

Accepted Answer

SkillzBench is pre-launch and actively being built — the waitlist is live now. The pricing hypothesis is a generous free tier, a Pro tier around $9-15/mo for more agents and unlimited enabled skills, a Teams tier around $15-25/seat/mo for org loadouts, allowlists and audit, and a Creator tier around $5-10/mo for verified authors, plus benchmark-as-a-service for priority runs.

Question 5

How is SkillzBench different from just browsing GitHub for agent skills?

Accepted Answer

GitHub gives you raw, unaudited repositories. SkillzBench continuously crawls and dedupes skills, pins every version to a commit SHA, security-scans and license-checks each one, benchmarks it in a sandbox to publish Lift, and then keeps tracking upstream changes with re-scans, re-benchmarks and notifications. As an independent lab, money never touches its scores — ratings and Lift stay separate and unbiased.

Custom Skill Development for Startup Teams