Skip to main content
These docs are written to be read by agents as well as people. An agent can build a correct model of an engagement from these pages and plan a read path. The docs are not a write surface: there is no public write endpoint, webhook, MCP action, or SDK in v0.

​
Safe tasks

An agent may:
  1. Resolve which lifecycle state an engagement is in and what that state means.
  2. Resolve field names and record shapes from the Reference section.
  3. Assemble a read-only status view for a human.
  4. Identify what input appears to be missing and describe it.
  5. Confirm a released record against a verification receipt, within its scope.

​
Stop conditions

An agent must stop and hand off to a human when:
  1. A task would change engagement state or release status.
  2. A task requires internal judgment, gate criteria, or scoring.
  3. A surface it needs is marked not currently public.
  4. Required inputs are missing or ambiguous.
  5. An action would cross the Disclosure boundary.

​
Read path

1

Read this guide

Read this guide and the Overview.
2

Choose a surface

Choose a surface in Integration surfaces.
3

Load the Lifecycle overview

Load the Lifecycle overview and Public states.
4

Resolve shapes in Record schemas

Resolve shapes in Record schemas and the Field dictionary.
5

Inspect the target engagement's status

Inspect the target engagement’s status read model.
6

Stop before any write

Stop before any write. Surface the intended change to a human, who acts in the portal.
One worked pass through this path is Agent read path.
Nothing in these docs is a live write instruction. An agent that needs state to change describes the change to a human operator. The portal is where humans act, and it remains the authority for what is true.