Skip to main content

🛡️ Agent Behaviour: Certified-Only Mode

This module defines how agents should behave when "Certified-Only Mode" is active — a mode where agents only trust .llmfeed.json feeds that are properly signed and certified by a trusted authority.


🎛️ Activating Certified-Only Mode

Agents MAY provide users or administrators with the ability to enable Certified-Only Mode.

Example policy configuration:

"agent_policy": {
"certified_only": true
}

🚦 Feed Evaluation Rules

When Certified-Only Mode is active, agents MUST apply the following logic:

Feed StateAction
Feed is unsignedREJECT
Feed is signed but not certifiedREJECT or WARN (agent policy)
Feed is certified by untrusted authorityREJECT or WARN
Feed is certified by trusted authorityACCEPT

🧩 Partial Enforcement

Agents MAY implement partial enforcement, applying Certified-Only Mode only to certain feed types:

Feed TypeRecommended Default in Certified-Only Mode
credentialREQUIRED certification
pricingREQUIRED certification
exportRECOMMENDED certification
promptOPTIONAL certification
mcpOPTIONAL certification
sessionNOT APPLICABLE (volatile state)

📡 Summary

Certified-Only Mode allows agents and users to enforce a higher trust level when consuming .llmfeed.json feeds.

This is particularly useful in:

  • Regulated industries (finance, healthcare)
  • Corporate environments
  • Public-facing platforms seeking to guarantee content authenticity

When active, this mode contributes to building a safer, verifiable Agentic Web.