diff --git a/decnet/profiler/behave_shell/_features/__init__.py b/decnet/profiler/behave_shell/_features/__init__.py
index 4e28b33d..13cd111e 100644
--- a/decnet/profiler/behave_shell/_features/__init__.py
+++ b/decnet/profiler/behave_shell/_features/__init__.py
@@ -25,6 +25,7 @@ from decnet.profiler.behave_shell._features.cognitive import (
     inter_command_latency_class,
 )
 from decnet.profiler.behave_shell._features.emotional_valence import (
+    arousal,
     valence,
 )
 from decnet.profiler.behave_shell._features.environmental import (
@@ -95,4 +96,5 @@ FEATURES: tuple[FeatureFn, ...] = (
     cleanup_behavior,
     multi_actor_indicators,
     valence,
+    arousal,
 )
diff --git a/decnet/profiler/behave_shell/_features/emotional_valence.py b/decnet/profiler/behave_shell/_features/emotional_valence.py
index 0c22928b..b255c88a 100644
--- a/decnet/profiler/behave_shell/_features/emotional_valence.py
+++ b/decnet/profiler/behave_shell/_features/emotional_valence.py
@@ -12,6 +12,7 @@ Step G.8: ``emotional_valence.frustration_venting`` (lands later).
 """
 from __future__ import annotations
 
+import statistics
 from typing import Iterator
 
 from decnet_behave_core.spec.envelope import Observation
@@ -19,6 +20,11 @@ from decnet_behave_core.spec.envelope import Observation
 from decnet.profiler.behave_shell._ctx import SessionContext
 from decnet.profiler.behave_shell._features._emit import make_observation
 from decnet.profiler.behave_shell._thresholds import (
+    AROUSAL_BANG_RUN_MIN,
+    AROUSAL_CALM_IAT_S,
+    AROUSAL_CAPS_RUN_MIN,
+    AROUSAL_FAST_IAT_S,
+    AROUSAL_MIN_IATS,
     EMOTIONAL_VALENCE_CONFIDENCE_CAP,
     VALENCE_FULL_CONFIDENCE_MIN,
     VALENCE_MIN_HITS,
@@ -63,3 +69,51 @@ def valence(ctx: SessionContext) -> Iterator[Observation]:
         value=value,
         confidence=_cap_soft(raw),
     )
+
+
+def arousal(ctx: SessionContext) -> Iterator[Observation]:
+    """Emit ``emotional_valence.arousal`` ∈ {low_calm, medium_engaged,
+    high_agitated}.
+
+    Three signals (any of which fires ``high_agitated``):
+
+    * ``ctx.caps_run_max ≥ AROUSAL_CAPS_RUN_MIN`` (5) — capslock rant.
+    * ``ctx.bang_run_max ≥ AROUSAL_BANG_RUN_MIN`` (3) — repeated bangs.
+    * The fastest typing burst's median IAT < ``AROUSAL_FAST_IAT_S``
+      (0.06) over a burst of ≥ ``AROUSAL_MIN_IATS`` (30) IATs.
+
+    ``low_calm`` — slowest qualifying burst's median IAT >
+    ``AROUSAL_CALM_IAT_S`` (0.30).
+
+    ``medium_engaged`` — fall-through.
+
+    Skip emission when no qualifying typing bursts. Confidence hard-
+    capped at 0.50; 0.30 below ``AROUSAL_MIN_IATS`` total typed IATs.
+    """
+    qualifying = [b for b in ctx.typing_bursts if len(b) >= 3]
+    if not qualifying:
+        return
+    fastest_med = min(statistics.median(b) for b in qualifying)
+    slowest_med = max(statistics.median(b) for b in qualifying)
+    total_iats = sum(len(b) for b in qualifying)
+
+    if (
+        ctx.caps_run_max >= AROUSAL_CAPS_RUN_MIN
+        or ctx.bang_run_max >= AROUSAL_BANG_RUN_MIN
+        or (
+            total_iats >= AROUSAL_MIN_IATS
+            and fastest_med < AROUSAL_FAST_IAT_S
+        )
+    ):
+        value = "high_agitated"
+    elif total_iats >= AROUSAL_MIN_IATS and slowest_med > AROUSAL_CALM_IAT_S:
+        value = "low_calm"
+    else:
+        value = "medium_engaged"
+    raw = 0.50 if total_iats >= AROUSAL_MIN_IATS else 0.30
+    yield make_observation(
+        ctx,
+        primitive="emotional_valence.arousal",
+        value=value,
+        confidence=_cap_soft(raw),
+    )
diff --git a/tests/profiler/behave_shell/test_emotional_valence_arousal.py b/tests/profiler/behave_shell/test_emotional_valence_arousal.py
new file mode 100644
index 00000000..50ab1f20
--- /dev/null
+++ b/tests/profiler/behave_shell/test_emotional_valence_arousal.py
@@ -0,0 +1,67 @@
+"""Step G.6: ``emotional_valence.arousal`` ∈ {low_calm, medium_engaged,
+high_agitated}.
+
+Hard 0.5 confidence cap.
+"""
+from __future__ import annotations
+
+from decnet.profiler.behave_shell import extract_session
+from decnet.profiler.behave_shell._parse import AsciinemaEvent
+
+
+PRIMITIVE = "emotional_valence.arousal"
+
+
+def _of(observations: list, primitive: str):
+    obs = [o for o in observations if o.primitive == primitive]
+    assert len(obs) == 1, f"expected exactly one {primitive}, got {len(obs)}"
+    return obs[0]
+
+
+def _typed(text: str, t0: float = 0.0, dt: float = 0.05) -> list[AsciinemaEvent]:
+    return [(t0 + i * dt, "i", c) for i, c in enumerate(text)]
+
+
+def test_no_typing_bursts_no_emission() -> None:
+    out = list(extract_session([(0.0, "i", "x")], sid="g6-empty"))
+    assert [o for o in out if o.primitive == PRIMITIVE] == []
+
+
+def test_high_agitated_via_caps_run() -> None:
+    """Long capslock streak fires high_agitated regardless of pace."""
+    text = "ls\rWHAT IS GOING ON HERE\rls\r"
+    obs = _of(list(extract_session(_typed(text), sid="g6-caps")), PRIMITIVE)
+    assert obs.value == "high_agitated"
+    assert obs.confidence <= 0.50
+
+
+def test_high_agitated_via_bangs() -> None:
+    text = "ls\rno!!!! something\rls\r"
+    obs = _of(list(extract_session(_typed(text), sid="g6-bang")), PRIMITIVE)
+    assert obs.value == "high_agitated"
+
+
+def test_high_agitated_via_fast_typing() -> None:
+    """Long fast burst (dt=0.04) fires high_agitated."""
+    text = "thequickbrownfoxjumpsoverthelazydog" * 2
+    obs = _of(
+        list(extract_session(_typed(text, dt=0.04), sid="g6-fast")), PRIMITIVE
+    )
+    assert obs.value == "high_agitated"
+
+
+def test_low_calm_slow_typing() -> None:
+    """Long slow burst (dt=0.40) fires low_calm."""
+    text = "thequickbrownfoxjumpsoverthelazydog" * 2
+    obs = _of(
+        list(extract_session(_typed(text, dt=0.40), sid="g6-calm")), PRIMITIVE
+    )
+    assert obs.value == "low_calm"
+
+
+def test_confidence_capped_at_05() -> None:
+    text = "thequickbrownfoxjumpsoverthelazydog" * 2
+    obs = _of(
+        list(extract_session(_typed(text, dt=0.10), sid="g6-cap")), PRIMITIVE
+    )
+    assert obs.confidence <= 0.50