Question 1

How does SkillzBench handle AI agent security vulnerabilities in community SKILL.md skills?

Accepted Answer

SkillzBench's Verify pillar security-scans every skill and pins each version to a commit SHA, so you always know exactly what code you're equipping. Each skill also receives a 0-100 Trust score built from security, quality, license, maintenance, and community signals. This matters because research found roughly of community-contributed skills carry at least one vulnerability.

Question 2

Does SkillzBench work with Claude Code, Codex, Cursor, and Gemini CLI agents?

Accepted Answer

Yes. SkillzBench is built around the SKILL.md instruction-package ecosystem used by Claude Code, Codex, Cursor, Gemini CLI, and chat-app agents. The Equip pillar delivers one MCP server plus per-agent loadouts, so you can toggle skills and tools individually for each agent and the agent only sees what was chosen.

Question 3

What's the difference between SkillzBench's Lift metric and a skill rating?

Accepted Answer

Lift measures how much a skill actually improves output: every skill is tested in a sandbox with the same model, with and without the skill, and the improvement is published in percentage points. Lift is always shown separately from Trust scores, and money never touches scores — so a paid placement can never masquerade as a benchmark result.

Question 4

Can SkillzBench help if I'm building skills for AI agents myself?

Accepted Answer

Yes — SkillzBench is planning a Creator tier (around $5-10/mo) for verified authors, plus benchmark-as-a-service for priority runs. Your skills go through the same verification, security scanning, and benchmarking pipeline, giving users independent proof that your work delivers real Lift.

Question 5

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

Accepted Answer

GitHub gives you raw repos with no verification, no benchmarks, and no way to compare. SkillzBench continuously crawls and dedupes skills, validates and license-checks them, benchmarks them in a sandbox, and lets you equip curated per-agent loadouts through one MCP server — then tracks upstream changes and re-benchmarks so your loadouts never silently break.

Curated Agent Loadouts for Founders