Trust & certification

Skills should be auditable before agents use them.

Mint A Skill treats generated capabilities as operational artifacts that need evidence, safety checks, and provenance.

Prompt-injection scanningFlags unsafe attempts to control the compiler or generated runtime behavior.

Rights-aware assessmentCaptures the declared rights basis and intended use before skill output.

Required validationValidation and certification stay part of the creation flow before saving.

Evidence-linked skillsOutputs can preserve source evidence, extracted steps, and provenance metadata.

Confidence scoringGenerated capabilities include scoring signals for fidelity, density, and trust.