Question 1

How does SkillzBench measure whether a SKILL.md package actually improves my agent?

Accepted Answer

SkillzBench is building a Benchmark pillar that tests every skill in a sandbox using the same model, with and without the skill, and publishes the resulting Lift in percentage points. Lift is always shown separately from Trust scores so performance and safety signals never get conflated.

Question 2

Does SkillzBench support script-bearing skills or only instruction-only SKILL.md packages?

Accepted Answer

Launch scope covers instruction-only skills — SKILL.md plus text references. Script-bearing skills will be listed link-out only, so you get discovery and pointers without SkillzBench executing unvetted scripts.

Question 3

How does SkillzBench handle agent skill security vulnerabilities like prompt injection?

Accepted Answer

The Verify pillar security-scans every skill version, pins it to a commit SHA, and runs license checks and quality judgments. Combined with a 0-100 Trust score built from security, quality, license, maintenance, and community signals, you can assess risk before equipping anything.

Question 4

When can I start using SkillzBench, and what will onboarding look like?

Accepted Answer

SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live now. Once public, onboarding centers on connecting one MCP server and configuring per-agent loadouts by toggling skills and tools per agent.

Question 5

Why use SkillzBench instead of just pulling skills straight from GitHub?

Accepted Answer

GitHub gives you raw, unpinned files with no security scanning, no benchmarked Lift, and no deduplication. SkillzBench continuously crawls and dedupes skills, verifies each version against its commit SHA, benchmarks it with and without the skill, and lets you equip curated per-agent loadouts through one MCP server — with upstream change tracking and re-benchmarking built in.

Curated Agent Loadouts for Researchers