fix(scene): spec the off-list fallacy boundary — decline or redirect, never improvise (#11)

The fallacy track's five lenses each pair a defect test with a
reverse-guard; a name outside the five (slippery slope, tu quoque /
whataboutism, no true scotsman, motte-and-bailey) had no specified
behavior, leaving the coach to improvise a ruling with neither — the
exact mislabeling risk the lens design exists to prevent.

The five lenses are now the complete ruling surface. An off-list name
gets, in order: the nearest covered lens offered as an explicit swap;
a redirect to the manipulation-recognition domain when the name lives
in shared/manipulation-taxonomy.md (whataboutism is the common
arrival); otherwise a plain decline — the insufficient_context
discipline generalized, since forcing a verdict without a defect test
is itself the reasoning error.

Three new invariant needles lock the boundary sentences in
modes/scene.md and shared/structures.md; the mutation loop covers them
automatically.

Issue #11 part (a); part (b) — growing the lens set one
defect-test+reverse-guard pair at a time — remains open.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ff8r7ajo54cRPm8v5MU4qr

Author: Imbad0202

modes/scene.md

@@ -180,6 +180,24 @@ The five lenses are in `shared/structures.md` (Fallacy-Recognition Lenses):
 `fallacy_false_dilemma`, `fallacy_ad_hominem`, `fallacy_strawman`,
 `fallacy_appeal`, `fallacy_equivocation`. One lens per round.
 
+**Off-list fallacy names.** The five lenses are the complete ruling surface of
+this track. When the user names a fallacy outside them (slippery slope, tu
+quoque / whataboutism, no true scotsman, motte-and-bailey, …), the coach says
+plainly that the lens set does not cover that name.
+It never improvises a ruling on it — a ruling without a paired defect test
+and reverse-guard is exactly the mislabeling risk the five-lens design exists
+to prevent. Then, in order:
+- If one of the five lenses genuinely applies to the same passage, offer that
+  lens by name and run the round on it — offered as a swap, never silently
+  substituted for what the user asked.
+- If the name is a manipulation technique in `shared/manipulation-taxonomy.md`
+  (whataboutism is the common arrival), say so and point the user to the
+  manipulation-recognition domain — recognition practice there, no fallacy
+  ruling here.
+- Otherwise decline to rule. The `insufficient_context` discipline generalizes:
+  forcing a verdict without a defect test is itself the reasoning error. The
+  passage remains workable in the frame palette as a new round.
+
 **Three rulings, never two.** Every fallacy call returns one of
 `fallacy` / `not_fallacy` / `insufficient_context`:
 - `fallacy` — the form commits the named fallacy.

scripts/check_invariants.py

@@ -209,6 +209,13 @@
      "fallacy recognition (is this argument a fallacy"),
     ("modes/scene.md", "fallacy-track-entry",
      "what selects the fallacy-recognition track"),
+    ("modes/scene.md", "fallacy-track-off-list-no-improvised",
+     "never improvises a ruling on it", "Fallacy-Recognition Track"),
+    ("modes/scene.md", "fallacy-track-off-list-decline",
+     "forcing a verdict without a defect test is itself the reasoning error",
+     "Fallacy-Recognition Track"),
+    ("shared/structures.md", "lens-set-complete-ruling-surface",
+     "the complete ruling surface", "Fallacy-Recognition Lenses"),
     ("passport/SCHEMA.md", "scene-fallacies-examined", "fallacies_examined"),
     ("platforms/claude-ai/passport/SCHEMA.md", "claude-ai-fallacies-examined",
      "fallacies_examined"),

shared/structures.md

@@ -90,7 +90,10 @@ Scene mode must cycle through all six lenses across each scene.
 
 ## Fallacy-Recognition Lenses
 
-Scene's **fallacy-recognition track** (modes/scene.md) uses these five lenses.
+Scene's **fallacy-recognition track** (modes/scene.md) uses these five lenses,
+and the five are the complete ruling surface — a fallacy named outside them is
+declined or redirected, never improvised (modes/scene.md, Off-list fallacy
+names).
 They are NOT frames — frames are interpretive and never ranked (redline 1);
 fallacy lenses adjudicate the *form* of an argument and DO return a ruling
 (`fallacy` / `not_fallacy` / `insufficient_context`). Each lens carries a