Portfolio layered authority (not flat)

Machine bundle: ai.json → portfolio_model (layers, Agent Blu surfaces, P0–P4, artifact homes). This file is the one-page human map; do not duplicate long tables into plans/ as a second SSOT.

Alignment rule (single sentence)

ContextControl.ai is the primary product objective. ContractPlane.ai is the control plane built alongside it. Dragonfly is the internal gate engine. OSSA and DUADP are the standards foundation. Agent Blu is the shared supervisory capability across all layers—not a standalone product line.

Architectural layers (0–4)

Layer	Product / plane	Role (one line)
0	OSSA	Protocol, spec, schemas, manifests, validators, release discipline
1	DUADP	Identity, discovery, addressing, execution contract, federation semantics
2	ContractPlane.ai	Control plane: ingress, trace propagation, tenant/workspace/project context, dispatch, approvals, policy gating, callbacks
3	Dragonfly (dragonfly.drupl.ai)	Internal only: CI acceleration, validation, preflight gates, audit automation, promotion readiness—not a distraction product before ContextControl stabilizes
4	ContextControl.ai	Primary user-facing product: workflows, Drupal / Tool API / governance integration surfaces

Agent Blu (cross-cutting)

One logical supervisory capability; multiple implementation surfaces (not one catch-all repo folder):

Surface	Where Blu shows up
Spec / pattern	OSSA reference manifests and agent patterns
Identity	DUADP discoverable and resolvable agent identity
Orchestration	ContractPlane supervisory orchestrator
Validation	Dragonfly policy / promotion consumer
Execution-facing	ContextControl.ai primary assistant / orchestrator UX

Canonical durable home for Blu artifacts (manifests, profiles, subagents, reusable definitions): PROJECTS/platform-agents/.agents/ (see existing @ossa Agent Blu assets). Skills / tool packs / registries: marketplace / ai-marketplace paths per BLU-OWNERSHIP.md. Policy bundles: cedar-policies (Blu + ContractPlane + ContextControl profiles). Elsewhere = reference, generated, or sprint-scoped—not a second canonical pile.

Operational priorities (P0–P4)

Authoritative list: ai.json → portfolio_model → operational_priorities. Short form: P0 execution spine → P1 ContextControl surface → P2 ContractPlane in tandem → P3 Dragonfly hardening → P4 OSSA/DUADP publish when stable.

What to stop

• Treating plans/ as a home for deployable code or infra (promote to owning repo + MR).
• Duplicate Blu “master plans” across random folders—one durable home under platform-agents (+ policy repo / marketplace as above).
• Mixing evergreen architecture with sprint notes without promotion (evergreen → GitLab Wiki per org policy; execution → plans/_ACTIVE/ or time-bounded paths).

Current state snapshot (2026-04-21)

• OSSA v0.5.1 released — new release publish pending.
• DUADP released — new release publish pending.
• ContextControl.ai openclaw Drupal module committed on release/v0.1.x: openclaw:sync_content Tool plugin, openclaw:test_connection, OpenClawRule config entity, OpenClawModelOwner modeler_api plugin. Push to GitLab pending (user action).
• Execution spine (P0): dual-gate model — Gate 1 api.copaw.us/execution/run, Gate 2 social-api.agentblu.ai/api/v1/invoke (SSE + Cedar). Gate 2 is PENDING / 502 blocker.
• blu-assistant canonical manifest in plans/blu-copaw/03-ossa-manifest/manifest.ossa.yaml — must migrate to platform-agents/agent-blu/ per canonical home rule above.
• 5 OSSA subagents created at ContextControl.ai/.agents/subagents/ — validated v0.5.1.

Handoffs (SoD + DRY)

Need	Open
SoD prose, conflict chain	BLU-OWNERSHIP.md
Domains, services, machine routing	ai.json
Workspace storage classes	BLU-WORKSPACE.md
Events / local outputs not canonical	BLU-EVENTS.md
Drupal CMS stack	BLU-DRUPALCMS.md
Vision “why” (thin)	../../plans/VISION.md — portfolio paragraph only; detail stays in ai.json

---

Update portfolio_model in ai.json when layers or P priorities change; keep this file a thin narrative mirror.