feat: bound AI-plan work to resource-control and stop zeroing missing tokens

Record the FIN-WP-0007 split after checking resource-control: they keep
intelligence resource identity and provision-level class I metering;
session tokens stay with State Hub; work-effectiveness is a reporting
join. Missing token counts are now unknown, not zero.

SCOPE.md

@@ -2,11 +2,18 @@
 
 ## In Scope
 
-- Budget, commitment, burn rate, runway projection, and token spend models
-- Manual CSV import for cloud and API costs (v0.1)
+- Budget, commitment, burn rate, runway projection, and booked token /
+  AI-plan spend
+- Manual CSV import for cloud, API, and subscription costs (v0.1)
 - Runway calculator with alert thresholds
 - Client-level cost attribution, engagement revenue and margin reporting, and
   billing-basis exports for external bookkeeping/invoicing systems
+- AI-plan entitlements, joins to State Hub session-token evidence, reporting
+  allocation of booked plan cost onto work, and effectiveness overlays
+  (work per token and work per euro) that never become booked spend
+- Consumption of resource-control planning evidence (forecasts, technical
+  usage, allocations, optimization, settlement statements) without taking
+  authority for those records
 - FOS §9 cross-hub signals: fin→dev (budget pressure), fin→ops (cost attribution), fin→canon (viability alerts)
 - hub-core generic primitives (domains, repos, messages, progress)
 
@@ -15,4 +22,13 @@
 - Legal invoice generation or numbering, bookkeeping, payment execution or
   tracking, and banking integration
 - Tax filing or legal entity management
-- Authority for client identity or infrastructure resource identity
+- Authority for client identity or infrastructure / intelligence resource
+  identity (`resource:…` remains resource-control)
+- Provision-level class `I` token metering against a portfolio resource
+- Technical usage, utilization, and allocation-driver evidence for
+  inventoried resources (resource-control, sourced from platform telemetry)
+- Parsing coding-agent session files or storing transcripts (State Hub /
+  llm-connect)
+- Inventing a `resource_id` for an AI plan that resource-control has not
+  published
+- A canonical token↔hour or token↔euro exchange rate