feat(context): reasoning-thread continuity, L3 doc filtering, and gate hardening

Threads reasoning_content across inference calls and journal renders, filters
markdown noise out of L3 repo-knowledge retrieval, adds PlanLinter H3 checks,
and tightens filesystem tool output/dir-listing behavior surfaced by prior
live QA (see project_readloop_campaign memory).

apps/tui-go/internal/app/server.go

@@ -153,16 +153,34 @@ func (m *Model) applyServer(msg protocol.ServerMessage) {
 			s.Tools = nil
 			s.StageTokensUsed = 0
 			s.addEvent(msg.OccurredAt, "StageStarted", msg.StageID)
+			for i := range s.PlanStages {
+				if s.PlanStages[i].ID == msg.StageID {
+					s.PlanStages[i].Status = PlanRunning
+					break
+				}
+			}
 		}
 	case protocol.TypeStageCompleted:
 		if s := m.session(msg.SessionID); s != nil {
 			s.CurrentStage = ""
 			s.addEvent(msg.OccurredAt, "StageCompleted", msg.StageID)
+			for i := range s.PlanStages {
+				if s.PlanStages[i].ID == msg.StageID {
+					s.PlanStages[i].Status = PlanCompleted
+					break
+				}
+			}
 		}
 	case protocol.TypeStageFailed:
 		if s := m.session(msg.SessionID); s != nil {
 			s.CurrentStage = ""
 			s.addEvent(msg.OccurredAt, "StageFailed", msg.StageID)
+			for i := range s.PlanStages {
+				if s.PlanStages[i].ID == msg.StageID {
+					s.PlanStages[i].Status = PlanFailed
+					break
+				}
+			}
 		}
 	case protocol.TypeInferenceStarted:
 		if s := m.session(msg.SessionID); s != nil {
@@ -303,6 +321,11 @@ func (m *Model) applyServer(msg protocol.ServerMessage) {
 		if s := m.session(msg.SessionID); s != nil {
 			s.addEvent(nowMillis(), "PlanLocked", strings.Join(msg.StageIDs, " → "))
 			s.LastOutput = "plan locked: " + strings.Join(msg.StageIDs, " → ")
+			s.PlanGoal = msg.Content
+			s.PlanStages = make([]PlanStage, 0, len(msg.StageIDs))
+			for _, id := range msg.StageIDs {
+				s.PlanStages = append(s.PlanStages, PlanStage{ID: id, Status: PlanPending})
+			}
 		}
 	case protocol.TypeWorkspaceBound:
 		if s := m.session(msg.SessionID); s != nil {