feat(research): ablations, broadcast and self-model experiments, counterfactuals

A framework for testing functional properties of the architecture
(Part C). None of it claims consciousness; every measure has an
operational definition in research.Measures.

- Ablations (attention, memory retrieval, affect, metacognition,
  commitments, volition, global workspace, curiosity, association, goals,
  narrative, notifications) and workspace overrides (force a matching
  candidate in or out of admission) live in a fork's cognitive
  configuration. Original minds refuse them on create and update; only
  snapshot restores may carry them, so they are part of the configuration
  digest. Excluded candidates are recorded with fate "excluded".
- Self-model experiments override a fork's self-model (capabilities,
  identity, goals, knowledge gaps) without touching policy. When an
  action's effective policy disagrees with the self-model, cognition
  records a self_model_mismatch and corrects the entry; experiments report
  self-model accuracy and correction latency.
- Counterfactual experiments replay a mind's recorded stimuli from a
  snapshot with chosen events omitted; the report lists what differed in
  beliefs, wants, intentions, memories, attention and capability
  decisions, in a form introspection can cite.
- Experiments record a protocol (kind, independent variable, controlled
  variables, measures), per-variant volition, plan, downstream-by-category
  (events reachable from each stimulus through recorded links),
  introspection and self-model measures and event ranges; comparisons add
  want, intention and plan-capability overlap, measure ranges, a
  descriptive architecture share and a downstream delta; reproducibility
  is labelled against the previous run of the same definition.
- research.Agency computes want→intention, intention→action, abandonment,
  commitment fulfilment and conflict, blocked intentions, replanning,
  outcome success and introspection accuracy; MetacognitiveCalibration
  relates confidence and expressed uncertainty in replies to later belief
  revision, with Brier scores for each half of the period.
- The sandbox test harness gains volition, planning, agency and research
  but not the motivation hook, whose wall-clock decay would make replay
  comparisons timing-dependent.

internal/replay/replaytest/harness.go

@@ -9,9 +9,11 @@ import (
 
 	"github.com/jackc/pgx/v5/pgxpool"
 
+	"github.com/andydixon/noema/internal/agency"
 	"github.com/andydixon/noema/internal/beliefs"
 	"github.com/andydixon/noema/internal/brain"
 	"github.com/andydixon/noema/internal/capabilities"
+	"github.com/andydixon/noema/internal/capabilities/host"
 	"github.com/andydixon/noema/internal/cognition"
 	"github.com/andydixon/noema/internal/conversation"
 	"github.com/andydixon/noema/internal/epistemics"
@@ -23,10 +25,13 @@ import (
 	"github.com/andydixon/noema/internal/minds"
 	"github.com/andydixon/noema/internal/motivation"
 	"github.com/andydixon/noema/internal/observability"
+	"github.com/andydixon/noema/internal/planning"
 	"github.com/andydixon/noema/internal/predictions"
+	"github.com/andydixon/noema/internal/research"
 	"github.com/andydixon/noema/internal/scheduler"
 	"github.com/andydixon/noema/internal/snapshots"
 	"github.com/andydixon/noema/internal/storage"
+	"github.com/andydixon/noema/internal/volition"
 	"github.com/andydixon/noema/migrations"
 )
 
@@ -54,9 +59,23 @@ func NewHarness(t *testing.T) *Harness {
 	for _, c := range capabilities.Builtins() {
 		caps.Registry.Add(c)
 	}
-	reg := cognition.NewRegistry(cognition.Stores{Memory: mem, Beliefs: bs, Conversations: conversation.NewStore(pool), Graph: graph.NewStore(pool), Assertions: epistemics.NewStore(pool)})
+	for _, c := range host.Capabilities(host.Options{}) {
+		caps.Registry.Add(c)
+	}
+	caps.Registry.Freeze()
+	vol := volition.NewStore(pool)
+	plans := planning.NewStore(pool)
+	gs := goals.NewStore(pool)
+	ps := predictions.NewStore(pool)
+	exec := &planning.Executor{Plans: plans, Volition: vol, Caps: caps, Predictions: ps, Events: es, Minds: ms, Memory: mem, Log: log,
+		Planner: &planning.Planner{Registry: caps.Registry, Effect: planning.PolicyEffect(caps)}}
+	rs := research.NewStore(pool)
+	ag := &agency.Service{DB: pool, Caps: caps, Plans: plans, Volition: vol, Goals: gs, Beliefs: bs, Predictions: ps, Events: es}
+	reg := cognition.NewRegistry(cognition.Stores{Memory: mem, Beliefs: bs, Conversations: conversation.NewStore(pool), Graph: graph.NewStore(pool), Assertions: epistemics.NewStore(pool), Agency: ag, Research: rs})
+	// No motivation hook: affect decay depends on elapsed wall-clock time,
+	// which would make replay comparisons in these tests timing-dependent.
 	cfg := brain.Config{DB: pool, Events: es, Minds: ms, Registry: reg, Pool: scheduler.NewPool(2), Log: log, Metrics: observability.NewMetrics(),
-		Services: brain.Services{Events: es, Memory: mem, Beliefs: bs, Goals: goals.NewStore(pool), Motivation: motivation.NewStore(pool), Predictions: predictions.NewStore(pool), Capabilities: caps}}
+		Services: brain.Services{Events: es, Memory: mem, Beliefs: bs, Goals: gs, Motivation: motivation.NewStore(pool), Predictions: ps, Capabilities: caps, Volition: vol, Plans: plans, Executor: exec, Minds: ms}}
 	ident := &identity.Capturer{DB: pool, Minds: ms}
 	return &Harness{Pool: pool, Events: es, Minds: ms, Snaps: &snapshots.Store{DB: pool, Events: es, Minds: ms, Identity: ident}, Brain: cfg, Identity: ident}
 }