Mechanical consequence of the .gitattributes in the previous commit, done in one
deliberate pass rather than left to trickle out as surprise diffs whenever
someone next touches a vendored file.
74 files, all CRLF -> LF, 71 of them in StarterAssets. `git diff -w` over this
commit is empty: nothing but line endings changed. DialogueTest.unity is
correctly untouched, being marked binary.
Co-Authored-By: Claude Opus 5 <[email protected]>
Scenes, prefabs and assets are now marked for UnityYAMLMerge, which is what
makes them mergeable at all — without it a scene conflict is resolved by a
line-oriented text merge that produces a file Unity may still load and that is
quietly wrong. The driver needs a one-time git config per machine; CLAUDE.md
carries the command.
Line endings are pinned to LF for every text type rather than left to
`text=auto`, so the 71 CRLF files vendored in StarterAssets stop being a source
of phantom diffs.
One exception, and it matters: DialogueTest.unity is marked `binary`. It is a
Unity binary SerializedFile rather than YAML, and it contains 128 lone CR bytes
— under `*.unity text eol=lf` git would rewrite those as line endings and
corrupt the scene. The rule carries the verification command and comes out as
soon as the file is genuinely text.
LFS is deliberately not enabled: the remote is self-hosted and its LFS support
is unverified, and enabling the filter against a server without it breaks
pushing. The rules to add later, and the migration step that has to accompany
them, are recorded in the file.
Co-Authored-By: Claude Opus 5 <[email protected]>
CLAUDE.md gains the writing/ layout, the voice-sheet pipeline and its direction
of truth, the sync commands, and a pointer that Ground Truth must never be
copied into this repo. The writer-owned/code-owned table now covers writing/.
The restructure plan records Phase 2 as executed, and closes its first open
question: the writing handbook does exist, in the Obsidian vault, so STYLE.md
became a digest with pointers rather than the reconstruction the plan assumed.
Also noted there: the vault's opening-scene spec records SC-101's tone check as
BLOCKED on a one-page Voice Sheet that was never written. STYLE.md now holds
that sheet with four fields explicitly unset.
Co-Authored-By: Claude Opus 5 <[email protected]>
Dialogue was already plain text, so the gap was never the .yarn files. It was
that the character material a writing session needs was either missing or stored
somewhere a writer could not comfortably work.
Three things were wrong. The writing handbook was not in the repo and was cited
by SC101.yarn as though it were. The best voice material — register, verbal tics,
success/failure/passive samples — was a \n-escaped string inside a JSON field.
The named NPCs had no sheets at all, existing only as their Yarn lines.
writing/ now holds the dialogue-facing layer, at the repo root so it gets no
.meta files and is not a Unity asset:
STYLE.md register, the five tone moves, the ban list, branching
shapes, the Wrong Truth rule, flag families, candidate rules
YARN-PRIMER.md the one page of syntax needed to write a scene
GLOSSARY.md in-world and project vocabulary
voices/ the eleven skill voices, one readable sheet each
characters/ Bartender, Chevalier Cassian Thal, Bradford Kane
scenes/ SC-101's brief: intent, beats, checks, constraints
lore/ the three identity readings and the rules that bind them
STYLE.md is a digest of the Mystery Writing Handbook in the Obsidian vault, with
pointers rather than copies. The vault stays the authoring home for design work,
and Ground Truth — the solution document — deliberately stays out of this repo.
The voice sheets are generated from skill_bible.json but are the source of truth
for the prose. tools/writing/sync_voices.py folds edits back; voicelib.py
guarantees notes -> dict -> notes is identity for all eleven, so --to-json on
unchanged sheets is byte-identical. VoiceSheetSyncTests fails the build on drift
in either direction, and was verified to fail on real drift rather than merely
to pass.
SC101.yarn's header is trimmed from prose rationale to the six rules that bind
while editing that file, with the reasoning moved to the scene brief.
EditMode 42/42. YarnCheck compiles 9 files / 35 nodes.
Co-Authored-By: Claude Opus 5 <[email protected]>
Both were test bugs; MotionInputMath itself is correct.
WorldToCameraRelativeInput_Yaw90 expected +1 for the camera-relative x. A camera
yawed +90 degrees faces +X, so world +Z is to its left and the value is -1. The
conversion applies Quaternion.Euler(0, -cameraYaw, 0) and Unity's Y rotation is
clockwise viewed from above, so forward maps to -right. The round-trip test
already pinned this convention and passes precisely because of it.
ScreenDeltaToWorldPan compared Vector3 with Assert.AreEqual, which is exact.
-10f * 0.02f lands one ulp from the -0.2f literal, so a correct result failed a
comparison that printed as identical at two decimals. Now compared per component
with a tolerance.
EditMode suite is 39/39.
Co-Authored-By: Claude Opus 5 <[email protected]>
CLAUDE.md documents repository layout, the Unity command line, the dialogue
content pipeline, validation, conventions, and known traps. AGENTS.md points at
it for non-Claude agents.
docs/restructure-plan.md covers both halves of the restructure: moving character
and voice prose out of JSON string fields into a writer-facing writing/ tree, and
migrating off the single-scene DialogueTest scaffold into Bootstrap/Systems/Level
scenes. Phase 0 is recorded as executed; everything from Phase 1 on is proposal.
Two findings in it were corrected after first being got wrong: the Unity CLI is
installed (just not on PATH for non-interactive shells) and does have build and
test commands, and the binary scene is not a stale save but a standing Unity
behaviour where Force Text converts only when the setting changes.
Co-Authored-By: Claude Opus 5 <[email protected]>
This is the scene the camera rig, UI layer, player control, and dialogue work in
the preceding commits were built against: 14 roots covering Room_101, the
Dialogue and Skill systems, the three UI canvases, the Cinemachine brain and
camera, the reveal volume, navigation, and the camera anchor.
Committed as-is, in binary, deliberately. The file is a valid Unity scene and
opens intact, but it is serialized as binary while the project is set to Force
Text, so it is not diffable or mergeable. Converting it needs the Asset
Serialization setting to be re-applied from the Editor UI, which cannot be done
from a batch-mode run — the setter does not trigger a reserialize pass. Checking
it in binary is strictly better than leaving days of scene wiring untracked;
the conversion is a follow-up, tracked in docs/restructure-plan.md.
Co-Authored-By: Claude Opus 5 <[email protected]>
DialogueInteractable, DialogueTransitionController, and DialogueUIVisibility are
reworked against the new UI layer: visibility is driven by UIStateController
rather than each component toggling canvases itself, which removes the
duplicated show/hide logic the two had accumulated.
NPCStandIn now ignores the interact button while UIStateController reports a
menu open, so pressing interact through an open panel no longer starts a
conversation behind it.
The yarnproject change is graph-editor node positions only.
Co-Authored-By: Claude Opus 5 <[email protected]>
Editor/PlayerControlSetup.cs is the one-shot menu that assembles the playable
player: the armature, the Cinemachine rig, the NavMesh agent, and the pointer
input wiring, so a scene can be brought up to a controllable state repeatably
rather than by hand.
StarterAssets input actions gain the pointer and interaction bindings the new
locomotion and interaction layers read, and PlayerArmature is updated to match.
ThirdPersonController blends the animator toward target speed scaled by input
magnitude. Previously a partial stick deflection or a click-to-move arrival
played a slowed run clip via MotionSpeed alone; scaling the blend makes those
use the walk cycle instead.
Co-Authored-By: Claude Opus 5 <[email protected]>
Runtime UI under Scripts/UI: a UIStateController owning menu/dialogue state, a
PlayerControlLock so opening a panel suppresses world input, panel visibility
helpers, and the menu input binding.
Views: the dialogue stream presenter and its options presenter and message row,
the portrait HUD, and the player menu with character, journal, and tab-bar
views. Skills gains SkillsPanelView and SkillRowView, which is why the Skills
asmdef now references UnityEngine.UI.
Prefabs/UI holds the authored canvases (dialogue stream, HUD, player menu) and
their rows. UILayerBootstrap remains a hand-added runtime fallback for scenes
where those prefabs are not instantiated; it deliberately does not auto-spawn.
Editor/UILayerSetup.cs is the one-shot menu that builds and wires those prefabs
into a scene, with UILayerSetupAutoRun reminding after a script reload.
Art/UI/Placeholder carries the placeholder portrait the HUD falls back to.
Co-Authored-By: Claude Opus 5 <[email protected]>
Camera: Cinemachine-based rig with an anchor, a state director, point-of-
interest framing, pan input, and dialogue framing volumes.
Player: click-to-move over the NavMesh, a pointer input layer, and a motion
arbiter that reconciles direct stick input with navigation-driven movement.
The pure math is isolated in NightclubArcadia.Locomotion.Math with an EditMode
suite, so the arbitration rules are testable without a scene.
Interaction: IClickInteractable plus a bridge that routes pointer clicks to it.
Supporting project settings: adds the Cinemachine package, the Interactable and
Ground tags the raycasts depend on, and NavMesh agent dimensions matched to the
player capsule (radius 0.3, height 1.8, climb 0.25).
Co-Authored-By: Claude Opus 5 <[email protected]>
lennartandClaude Opus 5