Skip to content

Validation

Validation is the process of checking an agent card against the A2A protocol specification and CapiscIO's trust requirements. It answers the question: "Is this agent correctly described and ready for production?"


What Gets Validated

The CapiscIO CLI validates agent cards across seven categories:

Category What It Checks
Schema Required fields, types, structure per A2A spec
Discovery Whether the card can be fetched from /.well-known/agent.json
Network HTTPS, response times, status codes
Version Compatibility Feature usage matches declared A2A protocol version
Security Authentication schemes, key presence
Endpoint Availability Whether declared URLs respond (optional)
Registry Readiness Additional fields needed for CapiscIO registration

Validation Modes

Progressive (Default)

Reports all findings without stopping at the first error. Suitable for iterative development — you see everything that needs fixing in one pass.

Strict

Treats warnings as errors. Required for production deployments and CI/CD pipelines where partial compliance is unacceptable.


Trust Relationship

Validation is the first step in the CapiscIO trust pipeline:

Validate → Register → Badge → Enforce

An agent that fails validation cannot register with the CapiscIO server and therefore cannot obtain a trust badge. Higher trust levels (2+) require stricter validation results.


When to Validate

  • During development — catch schema errors before deployment
  • In CI/CD — gate deployments on validation passing (--strict)
  • At registration — the server re-validates before accepting an agent
  • Periodically — detect drift as agent capabilities change

See Also