Skip to content

Clause 1 — Scope

1. Scope

(Normative)

1.1 Subject Matter

This specification defines the BPM/Agent Stack, an execution governance layer for AI agent architectures. It specifies the structural elements required for governed agent execution: per-activity governance attributes, process structure elements, governance infrastructure, and the stitching mechanism connecting execution governance to the Intent Stack’s governance context.

This specification formally governs three execution governance concerns:

a) Orchestration — how multiple agents SHALL be coordinated to execute authorized specifications, including swimlane-based responsibility assignment, subprocess-based governed decomposition, and message flow-based inter-agent communication;

b) Integration — how governed agents SHALL connect to the systems they need, including typed system attributes on governed activities, governed tool/API/MCP access, and governance context propagation through integrations;

c) Execution — how authorized work SHALL be performed within governing constraints, including process instantiation with assigned participants, live state management, and governance-quality audit trails.

1.2 What This Specification Covers

This specification covers:

a) The three-layer governance architecture and the complementary scope of each layer (Clause 5);

b) Agent deployment patterns as governance configurations, mapping operational species to structural governance properties (Clause 6);

c) The stitching mechanism connecting Intent Stack governance context to BPM/Agent Stack execution structure, including the bidirectional interface between the two specifications (Clause 7);

d) The governed activity model with 21 typed attributes derived from established frameworks (Clause 8);

e) Process structure elements for governed agent execution, per BPMN 2.0 (Clause 9);

f) Governance infrastructure spanning all processes within an execution domain (Clause 10);

g) Context, Memory, and Intent as structurally distinct concepts (Clause 11);

h) Validation mechanisms including auto-research patterns for governance boundary testing (Clause 12);

i) Self-referential governance — applying the specification to govern its own public/private boundary (Clause 13).

1.3 What This Specification Does Not Cover

This specification does not cover:

a) The content of governance at delegation interfaces — that is the scope of the Intent Stack (intentstack.org);

b) The model’s training-time character — that is the scope of Constitutional AI;

c) Implementation guidance for specific agent frameworks — that is the scope of framework-specific adaptation guides;

d) Organizational change management required for process-driven governance — that is addressed by the BPM CBOK Knowledge Area 6 (Business Process Transformation);

e) Mathematical foundations for execution governance — unlike the Intent Stack, which requires formal mathematical grounding for its structural claims, this specification derives authority from the existing formalization of BPMN 2.0, DMN 1.0, and the BPM CBOK.