v0.4.0: differentiator — REPL TDD + prose (daemon-side)

RED proofs (pre-v0.4.0):
- (getf (cognitive-verify ...) :gate-trace) → NIL (no trace)
- Reason suite: 12/0 (no gate-trace assertions)
- TUI actuator: no enrichment of rule-count/foveal-id

GREEN proofs (v0.4.0):
- gate-trace: ((:GATE mock-gate :RESULT :PASSED)), length 1
- Reason suite: 15/0 (new gate-trace assertions)
- TUI actuator enriches :rule-count, :foveal-id in payload

Prose:
- core-loop-reason.org: Gate Trace section — explains that no
  competitor can ship this because none has deterministic gates
  to trace. 0 LLM tokens per gate.
- core-loop-act.org: TUI Differentiator Enrichment section —
  documents :rule-count (HITL pending count) and :foveal-id flow.

org/core-loop-act.org

@@ -88,6 +88,14 @@ Because a skill's deterministic gate runs during Reason, but between Reason and
                                  (format stream "~a" (frame-message action))
                                  (finish-output stream))))))
 #+end_src
+
+** TUI Differentiator Enrichment (v0.4.0)
+
+The TUI actuator is the last point in the pipeline before the response leaves the daemon. It enriches the action plist with fields that power the TUI's differentiator visualizations:
+
+- ~:rule-count~ = ~(hash-table-count *hitl-pending*)~ — the number of pending HITL actions. The user watches this counter tick as they teach the agent their preferences.
+- ~:foveal-id~ = the current foveal focus from the signal context — enables the TUI's focus map status line.
+- ~:gate-trace~ — already attached by ~cognitive-verify~, flows through the action plist unchanged.
 #+end_src
 
 ** Action Dispatch (action-dispatch)

org/core-loop-reason.org

@@ -257,6 +257,12 @@ The system prompt assembly order — identity, tools, context, logs, mandates
 
 The deterministic engine is the strict guard. It receives a proposed action from the probabilistic engine and runs it through every registered deterministic gate, sorted by priority.
 
+**Gate Trace (v0.4.0)**
+
+As part of v0.4.0's TUI differentiator visualizations, ~cognitive-verify~ now accumulates a ~:gate-trace~ — a list of ~(:gate <name> :result <:passed|:blocked|:approval>)~ entries — as each deterministic gate processes the action. The trace is prepended to the result plist via ~list*~ and flows through the pipeline to the TUI actuator, which transmits it to the client.
+
+This is Passepartout's permanent UX advantage: no competitor can ship a gate trace because none has deterministic gates to trace. Claude Code, OpenClaw, and Hermes Agent all use prompt-based guardrails where the safety decision is invisible. In Passepartout, the user sees exactly which nine safety gates ran, what each decided, and why — all at 0 LLM tokens.
+
 Skills register deterministic gates via ~defskill~ with the ~:deterministic~ keyword. Each gate is a function that receives (action context) and returns either:
 - A modified action (the gate approves or adjusts the proposal)
 - A LOG or EVENT plist (the gate rejects the proposal with a reason)