Question 1

How does SkillzBench measure whether an agent skill actually improves data-analysis tasks?

Accepted Answer

SkillzBench is building a Benchmark pillar that tests every skill in a sandbox using the same model, running tasks both with and without the skill, then publishing the Lift in percentage points. Because the model, environment, and tasks are held constant, the Lift reflects the skill's contribution rather than seed noise or scaffold variance. Lift is always shown separately from ratings so it can't be gamed.

Question 2

What does the 0-100 Trust score actually evaluate?

Accepted Answer

The Trust score combines security, quality, license, maintenance and community signals. Every skill version is pinned to a commit SHA and goes through validation, license checking, security scanning, and quality judging before it's scored. This matters because published research found roughly of skills carry at least one vulnerability.

Question 3

Can I control which skills each of my agents uses?

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 for it. That means your data-analysis agent in Claude Code, Codex, or Cursor can carry a different verified skill set than your other agents.

Question 4

Is SkillzBench available now, and what will it cost?

Accepted Answer

SkillzBench is pre-launch and actively being built; the waitlist and newsletter are live today. The pricing plan is a generous free tier, Pro at roughly $9-15/month for more agents and unlimited enabled skills, Teams at roughly $15-25 per seat per month for org loadouts, allowlists and audit, and a Creator tier at roughly $5-10/month for verified authors, plus benchmark-as-a-service for priority runs.

Question 5

How is this different from just picking skills from GitHub or a leaderboard?

Accepted Answer

GitHub gives you popularity, not proof — and leaderboards suffer from known validity problems, with severe issues found in 8 of 10 popular AI agent benchmarks. SkillzBench instead pins every version to a commit SHA, security-scans and license-checks it, benchmarks it in a controlled same-model sandbox, publishes Lift separately from ratings, and keeps re-scanning and re-benchmarking as upstream code changes. Money never touches scores.

Agent Skill Benchmarking for Data Analysts