diff --git a/cmd/routines.go b/cmd/routines.go
index 553d5bd..5938676 100644
--- a/cmd/routines.go
+++ b/cmd/routines.go
@@ -252,9 +252,17 @@ func renderOnePreset(w io.Writer, p Preset, headingMarker string) error {
 			name = fmt.Sprintf("%s (%s)", name, *ex.ExerciseNotes)
 		}
 		fmt.Fprintln(w, name)
+		// An exercise added to a routine but never given target sets/reps
+		// arrives as one (or more) all-zero placeholder sets. Rendering them
+		// as `0@0` / `0@+0` / `0:00` looks like nonsense data; suppress the
+		// set lines but keep the exercise name so the LLM-agent reader still
+		// sees that the exercise is part of the routine.
+		if allSetsAreZero(ex.SetsData) {
+			continue
+		}
 		var lines []string
 		for _, s := range ex.SetsData {
-			lines = append(lines, fitdownSetLine(ex.ExerciseTypes, s))
+			lines = append(lines, fitdownSetLine(ex.ExerciseTypes, s.InputOne, s.InputTwo))
 		}
 		// Compress consecutive identical lines into Nx... notation.
 		for i := 0; i < len(lines); {
@@ -273,30 +281,66 @@ func renderOnePreset(w io.Writer, p Preset, headingMarker string) error {
 	return nil
 }
 
-// fitdownSetLine is split out of workouts.printFitdown so routines can render
-// the same notation without duplicating the type switch. Keeping it here
-// rather than in workouts.go keeps the workouts file untouched, at the cost
-// of a tiny bit of duplication of the WR/BR/AB/WD/DD/ND mapping.
-func fitdownSetLine(exTypes string, s PresetSetData) string {
+// allSetsAreZero reports whether every set in the slice has both inputs == 0.
+// Used to detect placeholder sets in routines (an exercise added but never
+// given a target) and an empty `ND` time entry in either renderer.
+func allSetsAreZero(sets []PresetSetData) bool {
+	if len(sets) == 0 {
+		return true
+	}
+	for _, s := range sets {
+		a, _ := s.InputOne.Float64()
+		b, _ := s.InputTwo.Float64()
+		if a != 0 || b != 0 {
+			return false
+		}
+	}
+	return true
+}
+
+// fitdownSetLine renders one set line in fitdown notation. Shared by the
+// workouts and routines renderers — both receive the same `exerciseTypes`
+// tag from Liftoff and emit identical notation. The fitdown spec only
+// documents `${reps}@${poundage}` and `Nx` compression; everything else
+// (WR/BR/AB/WD/DD/ND mapping, the `+`/`-` weight prefixes, the bodyweight
+// simplification below) is our extension.
+//
+// Simplification: when a WR/BR/AB set's weight component is 0 — i.e. a
+// bodyweight rep with no extra load or assistance — we drop the `@…`
+// suffix and emit just the rep count. "5" is clearer than "5@+0".
+func fitdownSetLine(exTypes string, inputOne, inputTwo json.Number) string {
+	weightZero := func() bool {
+		v, _ := inputOne.Float64()
+		return v == 0
+	}
 	switch exTypes {
 	case "WR":
-		return fmt.Sprintf("%s@%s", s.InputTwo, s.InputOne)
+		if weightZero() {
+			return fmt.Sprintf("%s", inputTwo)
+		}
+		return fmt.Sprintf("%s@%s", inputTwo, inputOne)
 	case "AB":
-		return fmt.Sprintf("%s@-%s", s.InputTwo, s.InputOne)
+		if weightZero() {
+			return fmt.Sprintf("%s", inputTwo)
+		}
+		return fmt.Sprintf("%s@-%s", inputTwo, inputOne)
 	case "BR":
-		return fmt.Sprintf("%s@+%s", s.InputTwo, s.InputOne)
+		if weightZero() {
+			return fmt.Sprintf("%s", inputTwo)
+		}
+		return fmt.Sprintf("%s@+%s", inputTwo, inputOne)
 	case "WD":
-		km, _ := s.InputTwo.Float64()
-		return fmt.Sprintf("%slb %.3fmi", s.InputOne, km/1.60934)
+		km, _ := inputTwo.Float64()
+		return fmt.Sprintf("%slb %.3fmi", inputOne, km/1.60934)
 	case "DD":
-		secs, _ := s.InputTwo.Int64()
-		km, _ := s.InputOne.Float64()
+		secs, _ := inputTwo.Int64()
+		km, _ := inputOne.Float64()
 		return fmt.Sprintf("%.2fmi %d:%02d", km/1.60934, secs/60, secs%60)
 	case "ND":
-		secs, _ := s.InputTwo.Int64()
+		secs, _ := inputTwo.Int64()
 		return fmt.Sprintf("%d:%02d", secs/60, secs%60)
 	default:
-		return fmt.Sprintf("[%s] %s %s", exTypes, s.InputOne, s.InputTwo)
+		return fmt.Sprintf("[%s] %s %s", exTypes, inputOne, inputTwo)
 	}
 }
 
diff --git a/cmd/routines_test.go b/cmd/routines_test.go
index f10f852..310d84f 100644
--- a/cmd/routines_test.go
+++ b/cmd/routines_test.go
@@ -172,6 +172,49 @@ func TestRenderOnePreset_ShowUsesH1(t *testing.T) {
 	}
 }
 
+func TestRenderOnePreset_AllZeroSetsSuppressed(t *testing.T) {
+	// An exercise added to a routine but never given target sets/reps
+	// arrives as one or more zero/zero placeholder sets. The exercise
+	// name should still appear (the LLM-agent reader cares that it's
+	// part of the routine) but the bogus "0@0" / "0:00" set lines must
+	// not.
+	p := Preset{
+		Name: "Free Weights",
+		ExerciseData: []PresetExerciseData{
+			{
+				ExerciseName:  "Bench Press",
+				ExerciseTypes: "WR",
+				SetsData:      []PresetSetData{{InputOne: json.Number("100"), InputTwo: json.Number("5")}},
+			},
+			{
+				ExerciseName:  "Placeholder",
+				ExerciseTypes: "WR",
+				SetsData:      []PresetSetData{{InputOne: json.Number("0"), InputTwo: json.Number("0")}},
+			},
+			{
+				ExerciseName:  "Sci-fit Upper Body",
+				ExerciseTypes: "ND",
+				SetsData:      []PresetSetData{{InputOne: json.Number("0"), InputTwo: json.Number("0")}},
+			},
+		},
+	}
+	var buf bytes.Buffer
+	renderOnePreset(&buf, p, "##")
+	out := buf.String()
+	if !strings.Contains(out, "Bench Press\n5@100") {
+		t.Errorf("real exercise should render normally; got:\n%s", out)
+	}
+	if !strings.Contains(out, "Placeholder") {
+		t.Errorf("placeholder exercise name should still appear; got:\n%s", out)
+	}
+	if strings.Contains(out, "0@0") {
+		t.Errorf("zero/zero WR set should be suppressed; got:\n%s", out)
+	}
+	if strings.Contains(out, "0:00") {
+		t.Errorf("zero ND duration set should be suppressed; got:\n%s", out)
+	}
+}
+
 func TestPickPreset(t *testing.T) {
 	resp := &presetsResponse{
 		PresetsWithoutFolder: []Preset{
diff --git a/cmd/workouts.go b/cmd/workouts.go
index c813015..c4422ac 100644
--- a/cmd/workouts.go
+++ b/cmd/workouts.go
@@ -308,28 +308,7 @@ func renderWorkoutsFitdown(w io.Writer, posts []Post) error {
 
 			var lines []string
 			for _, s := range ex.SetsData {
-				var line string
-				switch ex.ExerciseTypes {
-				case "WR":
-					line = fmt.Sprintf("%s@%s", s.InputTwo, s.InputOne)
-				case "AB":
-					line = fmt.Sprintf("%s@-%s", s.InputTwo, s.InputOne)
-				case "BR":
-					line = fmt.Sprintf("%s@+%s", s.InputTwo, s.InputOne)
-				case "WD":
-					km, _ := s.InputTwo.Float64()
-					line = fmt.Sprintf("%slb %.3fmi", s.InputOne, km/1.60934)
-				case "DD":
-					secs, _ := s.InputTwo.Int64()
-					km, _ := s.InputOne.Float64()
-					line = fmt.Sprintf("%.2fmi %d:%02d", km/1.60934, secs/60, secs%60)
-				case "ND":
-					secs, _ := s.InputTwo.Int64()
-					line = fmt.Sprintf("%d:%02d", secs/60, secs%60)
-				default:
-					line = fmt.Sprintf("[%s] %s %s", ex.ExerciseTypes, s.InputOne, s.InputTwo)
-				}
-				lines = append(lines, line)
+				lines = append(lines, fitdownSetLine(ex.ExerciseTypes, s.InputOne, s.InputTwo))
 			}
 
 			// Compress consecutive identical lines into Nx... notation
diff --git a/cmd/workouts_test.go b/cmd/workouts_test.go
index 55a1d3b..cd670db 100644
--- a/cmd/workouts_test.go
+++ b/cmd/workouts_test.go
@@ -66,23 +66,28 @@ func TestRenderWorkoutsFitdown_ExerciseNoteInParens(t *testing.T) {
 	}
 }
 
-// Set-line notation hasn't changed in this PR but is the load-bearing
-// piece routines share with workouts (via the fitdownSetLine duplicate).
-// One representative case per ExerciseType so a regression in either
-// renderer surfaces in CI.
+// Representative cases per ExerciseType so a regression in the shared
+// fitdownSetLine surfaces in CI. The bodyweight cases (BR/AB with zero
+// weight component) verify the @+0/@-0 simplification — the spec is silent
+// on these so the rule lives in code, not docs.
 func TestRenderWorkoutsFitdown_SetNotation(t *testing.T) {
 	cases := []struct {
+		name     string
 		exType   string
 		one, two string
 		want     string
+		notWant  string // optional negative assertion
 	}{
-		{"WR", "100", "5", "5@100"},     // weight/reps
-		{"BR", "0", "10", "10@+0"},      // bodyweight reps
-		{"AB", "20", "10", "10@-20"},    // assisted bodyweight (minus)
-		{"ND", "0", "495", "8:15"},      // no-data duration
+		{"WR weighted", "WR", "100", "5", "5@100", ""},
+		{"WR zero weight simplifies", "WR", "0", "5", "\n5\n", "5@0"},
+		{"BR with extra weight", "BR", "12", "5", "5@+12", ""},
+		{"BR zero added simplifies", "BR", "0", "10", "\n10\n", "10@+0"},
+		{"AB with assistance", "AB", "20", "10", "10@-20", ""},
+		{"AB zero assistance simplifies", "AB", "0", "10", "\n10\n", "10@-0"},
+		{"ND duration", "ND", "0", "495", "8:15", ""},
 	}
 	for _, c := range cases {
-		t.Run(c.exType, func(t *testing.T) {
+		t.Run(c.name, func(t *testing.T) {
 			w := bareWorkout()
 			w.ExerciseData[0].ExerciseTypes = c.exType
 			w.ExerciseData[0].SetsData[0] = SetData{
@@ -91,8 +96,12 @@ func TestRenderWorkoutsFitdown_SetNotation(t *testing.T) {
 			}
 			var buf bytes.Buffer
 			renderWorkoutsFitdown(&buf, []Post{w})
-			if !strings.Contains(buf.String(), c.want) {
-				t.Errorf("%s: expected %q in output; got:\n%s", c.exType, c.want, buf.String())
+			out := buf.String()
+			if !strings.Contains(out, c.want) {
+				t.Errorf("expected %q in output; got:\n%s", c.want, out)
+			}
+			if c.notWant != "" && strings.Contains(out, c.notWant) {
+				t.Errorf("did NOT want %q in output (should have been simplified); got:\n%s", c.notWant, out)
 			}
 		})
 	}