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/minds/store.go

@@ -89,6 +89,9 @@ type CreateParams struct {
 	Config      *CognitiveConfig
 	Notes       string
 	CreatedBy   string
+	// Research allows ablations and workspace overrides; only snapshot
+	// restores (forks, imports of research forks) set it.
+	Research bool
 }
 
 // Create inserts a mind with personality version 1 and an initial self-model.
@@ -127,6 +130,12 @@ func (s *Store) Create(ctx context.Context, p CreateParams) (Mind, error) {
 		cfg = *p.Config
 	}
 	cfg.Normalise()
+	if cfg.ResearchSettings() && !p.Research {
+		return Mind{}, fmt.Errorf("%w: ablations and workspace overrides are only allowed on research forks, never on an original mind", ErrInvalid)
+	}
+	if err := cfg.ValidateResearch(); err != nil {
+		return Mind{}, fmt.Errorf("%w: %v", ErrInvalid, err)
+	}
 	m := Mind{ID: ids.New(), Slug: p.Slug, DisplayName: p.DisplayName, Status: StatusActive, Config: cfg, Notes: p.Notes,
 		PersonalityVersion: 1, Personality: pers, TemplateID: p.TemplateID, CreatedAt: time.Now(), UpdatedAt: time.Now()}
 	err := storage.Tx(ctx, s.db, func(tx pgx.Tx) error {
@@ -237,6 +246,12 @@ func (s *Store) Update(ctx context.Context, id ids.ID, p UpdateParams) (Mind, er
 	if p.Config != nil {
 		c := *p.Config
 		c.Normalise()
+		if c.ResearchSettings() && m.ForkedFromSnapshot == nil {
+			return Mind{}, fmt.Errorf("%w: ablations and workspace overrides are only allowed on forks, never on an original mind", ErrInvalid)
+		}
+		if err := c.ValidateResearch(); err != nil {
+			return Mind{}, fmt.Errorf("%w: %v", ErrInvalid, err)
+		}
 		m.Config = c
 	}
 	cfgJSON, _ := json.Marshal(m.Config)