Layer 2 of 12·Slice: Human and intent surface·No dedicated platform

Scope of this page. Reference material for the Goal & Intent layer: what it owns, what it must not own, and how Synedrix OS implements it today. It is descriptive, not normative — where this page and the Layer Taxonomy Table disagree, the taxonomy table wins.

What this layer is

Goal & Intent is the layer where a business objective becomes something the rest of the system can act on: a structured goal schema, an objective definition, a playbook, or a KPI target, rather than a paragraph of prose or a verbal instruction. Its job is translation, not execution — turning “what a human or business wants to happen” into a shape that Orchestration (Layer 3) can decompose into a task graph and route.

This translation step matters in a governed multi-agent system because everything downstream — task decomposition, agent invocation, policy evaluation — depends on intent being expressed consistently rather than being re-interpreted ad hoc at each layer. A goal that reaches Orchestration already structured is one Layer 3 can plan against; a goal left as free text would push interpretation work into whichever layer happened to receive it first, which is exactly the kind of layer-boundary leakage the taxonomy table’s ownership guidance rules out.

flowchart LR
  n1([Goal Intent]):::term
  subgraph What it owns
    n2[Goal schema]:::agent
    n3[Objective definitions]:::agent
    n4[Playbooks]:::agent
    n5[KPI targets]:::agent
  end
  n1 --> n2
  subgraph What it must not own
    n6[Domain-specific execution logic]:::warn
    n7[Domain Pack Layer]:::term
    n6 --> n7
    n8[Raw tool calls]:::warn
    n9[Tool Abstraction Layer]:::term
    n8 --> n9
  end
  n1 --> n6
  subgraph Interfaces with
    n10[Human interface]:::check
    n11[Orchestration]:::check
    n12[Governance]:::check
  end
  n1 --> n10

What it owns

Per the Layer Taxonomy Table, Goal & Intent owns the structured representations that a business objective is translated into — not the domain logic that later acts on them. Concretely, that means:

  • Goal schema
  • Objective definitions
  • Playbooks
  • KPI targets

What it must not own

Goal & Intent stops at translation: once an objective is structured, deciding how a specific domain will pursue it, or which tool call satisfies it, belongs elsewhere.

ConcernCorrect owner
Domain-specific execution logicDomain Pack Layer
Raw tool callsTool Abstraction Layer

Interfaces with

A structured goal is only useful once it reaches the layers that will act on it or review it, so Goal & Intent interfaces with three layers:

  • Human interface — objectives and KPI targets originate from a human through Layer 1’s surfaces before they are captured here in structured form.
  • Orchestration — a structured goal is the input Layer 3’s planner decomposes into a task graph; without this layer’s schema, Orchestration would have nothing consistent to plan against.
  • Governance — playbooks and KPI targets can themselves encode constraints (thresholds, required approvals) that Layer 9 must be able to evaluate against, so the two layers stay coordinated on what an objective is allowed to imply.

Current specification depth

Current specification depth. Goal & Intent is specified by three canonical documents. The Goal Model Spec holds authority for goal types, goal instances, lifecycle, identity, tenant and business-instance scoping, success criteria and terminal states. The Goal Planning Strategy holds authority for decomposing a goal, resolving capabilities and binding executors. The Task Graph Spec holds authority for the persisted work graph a goal becomes. The row in the Layer Taxonomy Table remains the definition of the layer’s boundary, echoed by README §3.1 and the consolidated master §4. Orchestration (Layer 3) still carries the dispatch machinery described in the Orchestration & Runtime Strategy, and any constraint a playbook or KPI target implies is evaluated by the Policy Guard Agent (os-agt-002) under the Agent Registry, not in a Goal & Intent spec. Playbooks and KPI targets as a topic still have no dedicated document.

OS agents at this layer

No OS-owned agent operates natively at this layer. Goal schemas, objective definitions, and playbooks are structured here but consumed by Layer 3’s orchestration and planning logic — most directly the Orchestrator Agent (os-agt-001), which decomposes a structured goal into a task graph. Per the Agent Registry, the twelve OS agents are assigned only to Layers 3, 4, 5, 7, 9, and 10.

Layer principles that apply

Two of the taxonomy table’s layer principles bear directly on this layer:

  • 1. Shared before specific — a goal schema or playbook structure that could serve any domain pack belongs here as a shared OS primitive, rather than being reinvented inside each domain pack that needs to express an objective.
  • 2. Controls before execution — a playbook or KPI target that implies an approval requirement must still pass through Governance & Controls before the objective it describes is acted on; structuring an intent is not the same as authorizing it.

Synedrix OS · Layer reference · Goal & Intent