Every skill is sealed before it touches a robot.
A two-sided marketplace where developers publish robot skills, knowledge, media and personalities — and every submission must clear the ThinkNEO two-layer validation sandbox before an operator can install it.
The validation seal is the moat
An enterprise will never install a skill from an unknown developer onto a robot inside its operation without a guarantee it is safe and in compliance. The seal turns “code from a dev I don’t know” into “a verified skill I can install on a robot inside my bank.”
Layer 1 — Binary security floor
The same minimum bar for every item, automated and cheap: prompt-injection scan, PII/secrets scan, provider allow-list enforcement, and package integrity. Fail any check and it goes back to the developer with a structured report.
- · Prompt-injection scan
- · PII / secrets scan
- · Provider allow-list enforcement
- · Package integrity / signature
Layer 2 — Risk-graded validation
Items that clear the floor are classified by criticality. Low risk seals automatically; medium adds policy + data-boundary tests; high risk (physical movement) requires human review before a seal is ever issued.
- · Low — automated, seal direct
- · Medium — extended tests + policy
- · High — physical movement → human review required
Four content types, one pipeline
Behavior / Skills
New abilities and routines, delivered OTA and sandboxed before full permission.
Media / Response
What the robot says or shows — governed inline on the output path.
Knowledge / Context
RAG knowledge bases queried at runtime — access governed and audited.
Personality
Custom persona configuration, including commissioned personality builds.