Add unique identifier for test_phase3c_walk_recovery.gd

docs/phase9_task4_refactor_spec.md

@@ -419,6 +419,37 @@ variant not referenced by the factory. The new `StickmanRig` script targets `mas
 | D8 | Rig enables its own mod stack in `_ready()` | Facing/bend are meaningless until the stack is live; the rig should self-enable at runtime (single consumer today always enables it). |
 | D9 | Rig exposes `get_bend_joint_global_position()` | Lets the harness drop `BEND_JOINT_BONE_PATHS`/`_bend_joint_bones` entirely; the rig owns the whole bend domain. |
 
+## 9a. Round N — whole-rig Y-axis mirror (2026-09-05 design change)
+
+**Decision (user-approved):** facing LEFT is now rendered as a **whole-rig Y-axis mirror** —
+`Master.scale.x = -1` (RIGHT/FORWARD → `(1,1)`) — replacing the per-part/head mirroring. This flips
+the head **and** body together so the figure faces the correct direction.
+
+- `_apply_head_flip()` and the `Body/Head.scale.x` mirror are **removed** (the Head Pivot node's
+  driver transform is untouched).
+- `PROFILE_FLAGS` (per-joint `flip_bend_direction`) and `Z_ORDER_BY_PROFILE` are **kept
+  provisionally** (unchanged). The mirror reflects the whole skeleton + `IK_Targets` + mounted
+  `Body/*` geometry, but does **not** affect depth (draw order); whether the bend flags can be
+  collapsed to a single canonical set must still be verified empirically (a root mirror is not
+  provably reflection-invariant for TwoBoneIK's `flip_bend_direction` sign).
+- **Walk-clip mapping — Option A (single canonical clip):** `walk_right` is the canonical walk.
+  For `FacingProfile.LEFT` the rig root is X-mirrored and the **same `walk_right`** clip plays
+  mirrored; `walk_left` is no longer used at runtime. The animation `.:facing_profile` tracks are
+  neutralized/removed — facing is set explicitly by `set_facing_profile()` / `walk_to()`.
+
+**Follow-up (implemented, tested):** two head-related fixes were required to make the LEFT root
+mirror render the head correctly. (1) The head `RemoteTransform2D` (`Skeleton2D/Torso/Head/Pivot`)
+no longer sets `update_scale = false` — it pushes the **full transform** like every other `Body`
+driver, so `Body/Head.scale` stays identity under the mirrored root (the old partial-channel push
+re-canonicalized the scale and caused per-frame Y-flips/wrap-jumps). (2) The `SkeletonModification2DLookAt`
+that aims the Head bone is **not mirror-invariant**: under the LEFT root mirror it writes a bone
+rotation 180° off the FORWARD aim, flipping the head to hang below the neck. New
+`_apply_head_lookat_mirror_mode()` (called from `_apply_profile()`) disables the LookAt and pins the
+head bone to the FORWARD canonical aim (π) when facing LEFT; `_pin_mirrored_head_rotation()`
+re-asserts the pin each `_physics_process` frame while ANIMATED/RECOVERING so recovery's stack
+re-arm can't let LookAt flip the bone. RIGHT/FORWARD re-enable the LookAt. Consequence: interactive
+head-aiming while facing LEFT is intentionally static.
+
 ## 10. Test plan
 
 1. **Parse check** (same as prior tasks):