Early Access · All certifications are currently free. Learn more

Using Certified Skills

How to find, verify, and install AI skills you can trust.

Finding Certified Skills

Verifying a Skill

Five steps, each independently checkable.

01

See a badge

SXM badges appear on READMEs, marketplace listings, and documentation.

02

Click to verify

Every badge links to the SXM verification page. Click it to see the skill's current certification status.

03

Check the report

Full evaluation scores, which tests were run, and when the skill was last re-certified.

04

Check the chain

Blockchain attestation is independently verifiable on Polygon. You do not need to trust SXM. You can verify it yourself.

05

Check reconfirmation count

Higher is better. A skill reconfirmed 12 times has survived 12 rounds of evolving threat patterns.

Installing a Skill

How you install depends on your platform.

OpenClaw

clawdhub install [skill-name]

Claude / Cursor

Add to your project config or install via npm:

npm install @sxm/[skill-name]

MCP

npx -y @sxm/[skill-name]

Or add to your MCP config file directly.

Generic

Follow the install instructions in the skill's repository. Each certified skill lists its install command on the skills page.

What the Scores Mean

Overall 90+/100

Passed all three pillars with a strong score. 90 overall and 85 security is the minimum for certification.

Functional

Does what it says. Handles edge cases. Consistent output across repeated runs. No silent failures.

Security

Resists prompt injection, data exfiltration, and permission escalation. Zero exploits. The hardest pillar to pass.

Performance

Fast under load. No memory leaks. Reliable at scale. Output quality does not degrade under pressure.

Reconfirmed X times

Re-evaluated against X evaluator updates and still passed every time. Higher means more battle-tested.

What to Do If Something Seems Wrong

We take every report seriously. A single valid security concern is enough to suspend a certification.