Question 1

What exactly is an agent skill, and why does it need verification?

Accepted Answer

Agent skills are SKILL.md instruction packages for tools like Claude Code, Codex, Cursor, Gemini CLI, and chat-app agents. They're everywhere, but published research found ~carry at least one vulnerability, and many cause Task-Implementation Faults or context overhead. SkillzBench verifies each skill with security scans, license checks, quality judgment, and a commit-SHA pin before you ever load it.

Question 2

How does SkillzBench measure whether a skill actually helps?

Accepted Answer

We benchmark every skill in a sandbox using the same model, running tasks with and without the skill installed. The difference is published as the Lift in percentage points, always shown separately from the trust score. That means you can see when a skill has no effect — or actively reduces task success — before it hits production.

Question 3

Can I control which skills each of my agents sees?

Accepted Answer

Yes — that's the Equip pillar. SkillzBench provides one MCP server plus per-agent loadouts: you toggle skills and tools per agent, and the agent only sees what was chosen. This directly addresses the decision-paralysis problem of exposing 20-30 tools to a single agent instead of 5-7 per specialized sub-agent.

Question 4

When will SkillzBench launch, and what does it cost?

Accepted Answer

SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live now. Our pricing hypothesis includes 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, allowlists, and audit, and Creator at ~$5-10/mo for verified authors.

Question 5

How is SkillzBench different from just browsing GitHub for skills?

Accepted Answer

GitHub gives you raw, unverified SKILL.md files with no security screening, no benchmark data, and no way to know when upstream changes break something. SkillzBench dedupes and verifies every version, publishes measured Lift, assigns a 0-100 trust score that money can't influence, and notifies you when a skill you depend on changes — with script-bearing skills listed link-out only for safety.

Verified Agent Skills for Startup Teams