Audio: FMOD Studio + Vercidium Audio system (phases 1-15) and Windows fixes

.agents/skills/plan-le/SKILL.md

@@ -0,0 +1,30 @@
+---
+name: plan-le
+description: LuxEngine implementation planning — a source-grounded, phased plan with a verified current-state ledger, user decisions, engine-fit analysis (systems, threads, ownership), independently verifiable phases, risks, and open questions. Use before multi-file or multi-session work, or when asked for a plan, design, or roadmap. Plans only.
+---
+
+# plan-le (Codex adapter)
+
+**The workflow body is `.claude/skills/plan-le/SKILL.md`. Read it and follow it.** This file exists
+only for Codex discovery and tool translation; it deliberately does not restate the workflow, so the
+two cannot drift.
+
+## Translation notes
+
+- `/plan-le` in shared docs means this skill, normally invoked as `$plan-le`.
+- `$ARGUMENTS` is the goal to plan. With no arguments, ask what to plan.
+- "Ask the user" and "plan mode" in the body describe intended actions. Use Codex's equivalent
+  (a direct question, or its planning/approval flow); where none exists, present the plan in chat
+  and wait for approval.
+- Claude-specific tool names describe an intended action. Use the equivalent Codex-native tool.
+- Step 4 (web research) needs a web search/fetch capability. If this Codex environment has none,
+  say so in the plan, mark the Research brief as not done, and list what should be researched.
+- Paths starting with `.claude/`, `.agents/`, `docs/`, or `scripts/` are relative to the repository
+  root.
+- `/dev`, `/cr`, `/send-pr`, and `/profile` in the body refer to the repository skills of the same
+  names.
+
+## Boundary
+
+`plan-le` does not edit engine code, commit, push, or open pull requests. Writing a
+`docs/<AREA>_PLAN.md` file is the only file it may create, and only when the user agrees.

.agents/skills/profile/SKILL.md

@@ -0,0 +1,27 @@
+---
+name: profile
+description: LuxEngine performance investigation — rule out the present/VSync ceiling, decide CPU- vs GPU-bound, capture Tracy/RenderDoc when needed, and report before/after numbers on a fixed protocol. Use for low FPS, hitches, slow loads, or optimization and benchmark requests.
+---
+
+# profile (Codex adapter)
+
+**The workflow body is `.claude/skills/profile/SKILL.md`. Read it and follow it.** This file exists
+only for Codex discovery and tool translation; it deliberately does not restate the workflow, so the
+two cannot drift.
+
+## Translation notes
+
+- `/profile` in shared docs means this skill, normally invoked as `$profile`.
+- `$ARGUMENTS` is the symptom or target supplied with the invocation. With no arguments, run the
+  triage in Step 1.
+- Claude-specific tool names in the body describe an intended action. Use the equivalent
+  Codex-native tool.
+- Paths starting with `.claude/`, `.agents/`, `docs/`, `scripts/`, or `tests/` are relative to the
+  repository root.
+- Steps that need the running editor (panels, Tracy connection, RenderDoc) require the user unless
+  you can launch and observe the app; say which measurements you could not take yourself.
+- `/cr` in the body refers to the `cr` repository skill.
+
+## Boundary
+
+`profile` does not commit, push, or open pull requests.

.agents/skills/send-pr/SKILL.md

@@ -12,7 +12,7 @@ deliberately does not restate the rules, so the two cannot drift.
 ## Translation notes
 
 - `/send-pr` in shared docs means this skill, normally invoked as `$send-pr`.
-- The rule list in the body (rules 1–19, tiered must-fix / should-fix / consider) is shared with
+- The rule list in the body (rules 1–20, tiered must-fix / should-fix / consider) is shared with
   `dev` and `cr`. It is maintained in one place; do not fork it here.
 - Claude-specific tool names describe an intended action; use the equivalent Codex-native tool and
   preserve every safety gate.

.agents/skills/shader-debug/SKILL.md

@@ -0,0 +1,27 @@
+---
+name: shader-debug
+description: LuxEngine shader debugging — compile errors, edits with no visible effect, black or garbage output, startup crashes after a shader change, binding collisions, and device-lost or vendor-specific GPU faults. Use whenever a .glsl/.glslh/.hlsl change misbehaves or rendering output is wrong.
+---
+
+# shader-debug (Codex adapter)
+
+**The workflow body is `.claude/skills/shader-debug/SKILL.md`. Read it and follow it.** This file
+exists only for Codex discovery and tool translation; it deliberately does not restate the workflow,
+so the two cannot drift.
+
+## Translation notes
+
+- `/shader-debug` in shared docs means this skill, normally invoked as `$shader-debug`.
+- `$ARGUMENTS` is the symptom supplied with the invocation; use it to pick the triage section.
+- Claude-specific tool names in the body describe an intended action. Use the equivalent
+  Codex-native tool.
+- Paths starting with `.claude/`, `.agents/`, `docs/`, `scripts/`, or `tests/` are relative to the
+  repository root. Shader and cache paths under `Resources/` are relative to the `Editor/` folder.
+- Steps that need the running editor (Ctrl+Shift+R, Renderer Debugger views, RenderDoc) require the
+  user unless you can launch and observe the app; the offline `glslc` / `spirv-val` path works
+  without it.
+- `/cr` in the body refers to the `cr` repository skill.
+
+## Boundary
+
+`shader-debug` does not commit, push, or open pull requests.