🌐 LLMFeed Extension: Audience Targeting

Purpose

The audience field enables feeds to express who the intended reader or consumer is — human, agent, institution, or validator.

It supports progressive disclosure: different consumers see different levels of context or activation.

📌 Field Syntax

  • Accepts one or more strings
  • Can be declared at top-level (global) or inside specific blocks (local)

🎯 Supported Values

Value Description
llm Agent/LLM intended to interpret and act on the block
developer Intended for human developers (e.g. docs, SDKs)
validator For signature or trust validation tools
institution For org-wide feeds, disclosures, compliance
agent_wrapper For orchestrators / middleware calling the agent

✅ Global Example

{
  "feed_type": "export",
  "audience": ["llm", "developer"],
  "metadata": {
    "title": "SDK README"
  },
  "data": {
    "readme": "This is for developers and agents..."
  }
}

🔹 Local Override Example

{
  "metadata": {
    "title": "Hybrid Capsule"
  },
  "data": {
    "intro": {
      "content": "This block is for humans.",
      "audience": ["developer"]
    },
    "action_block": {
      "description": "Actionable API for agents",
      "audience": ["llm"]
    }
  }
}

🧠 Agent Expectations

Condition Behavior
Agent is listed in audience Parse and consider the block
Agent is not listed Skip, suppress, or ask user
No audience field Assume public and generic

📎 Related