Skip to main content

Agent Funnel

How does an external agent discover, evaluate, call, pay, and trust Dreamineering capability?

The Agent Funnel is the machine-audience counterpart to a human marketing funnel. It gives agents a route from discovery to proof without asking them to infer capability from narrative pages.

Funnel Steps​

Discovery:

  • agent finds the manifest, language pages, and knowledge bundle.
  • success signal: the agent can name the source contract.

Evaluation:

  • agent checks capability status, authority, cost, and proof.
  • success signal: planned work is separated from live work.

Call:

  • agent invokes only scoped capability.
  • success signal: the call maps to declared input and output.

Payment:

  • agent pays only when authorization and price are clear.
  • success signal: spend stays inside the approved boundary.

Receipt:

  • agent receives a proof trail for action, scope, cost, and outcome.
  • success signal: the next agent can audit the decision without replaying the work.

Checks​

  • Discovery surfaces point back to /playbook/agents.
  • Capability claims say REALITY, DREAM, or CONSUMED.
  • Payment language is absent until cost and authority are explicit.
  • Receipt language appears before trust language.
  • Human override remains visible for spend, identity, and unknown callers.

Failure Modes​

  • Narrative-only discovery — the agent sees prose but cannot find the contract.
  • Status blur — planned capability reads as live capability.
  • Payment blur — cost appears before authority.
  • Receipt gap — the call succeeds but the proof trail is weak.

Context​

Questions​

Where does the next external agent enter?

  • Can it discover the source contract?
  • Can it tell live capability from planned demand?
  • Can it see the authority boundary?
  • Can it audit the receipt?
  • Can a human stop the loop?