菜单
Answer questions about orkid's DrawingInterface (DWI), 2D/3D quad rendering, line drawing, fullscreen quads, tiled quads, screen-space rendering helpers, and Python DWI bindings. Use when the user asks about 2D drawing, quad rendering, or screen-space primitives.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Orchestrate multi-agent engine work on orkid — decomposing milestone specs into lanes, assigning machines/branches, briefing implementer and gate-runner subagents, reviewing, adjudicating gates, and integrating. Use when coordinating subagent-based implementation of a spec/milestone rather than implementing directly.
How to author procedural content in orkid using HyperSyn (Reality Synthesizer) — a Python-as-DSL system for composing particle systems, 2D/3D procedural textures, mesh/SDF operations, and (eventually) audio/sequencer graphs into compact runtime experiences. Use when the user asks about HyperSyn, ptex2d, ptex3d, hypermesh, dataflow DSL, FXV2 codegen from dataflow, subprocess graph validation, or the hypergraph composition layer.
Operate THIS machine as an obtnet fleet node — launching the node daemon correctly, what the controller may do to this machine (jobs, builds into the staging, checkout gitsync/sync), node-side etiquette and troubleshooting. Use when setting up, running, or debugging the node side of the fleet (the controller-side skill is `obtnet`).
Drive the obtnet fleet (distributed build/run/sync across machines) via obt.net.py — node routing, async jobs, artifact fetch, git-free tree sync. Use whenever work should run on another machine or files must move between machines without git.
Answer questions about orkid's OBT build system, ork.build.py, CMake structure, module layout, ork.python environment, staging directories, pyext builds, and environment variables. Use when the user asks about building, compiling, CMake, or project structure.
Answer questions about orkid's ComputeInterface (CI), compute shader dispatch, indirect dispatch, SSBO binding for compute, memory barriers, image/sampler binding, dispatch phases. Use when the user asks about compute shaders, GPU compute, or dispatch.
| name | orklev2-dwi |
| description | Answer questions about orkid's DrawingInterface (DWI), 2D/3D quad rendering, line drawing, fullscreen quads, tiled quads, screen-space rendering helpers, and Python DWI bindings. Use when the user asks about 2D drawing, quad rendering, or screen-space primitives. |
| user-invocable | false |
When answering questions about 2D drawing, screen-space quads, or line rendering in orkid, consult the files below. All under ork.lev2/.
| Component | File |
|---|---|
| DWI Interface | inc/ork/lev2/gfx/dwi.h (lines 9–56) |
| Vulkan DWI | src/gfx/vulkan/headers/vulkan_ctx.h (lines 139–142) |
| Python Bindings | pyext/src/pyext_gfx.cpp (lines 298–322) |
The DrawingInterface provides methods for rendering 2D/3D quads and lines, primarily used for post-processing, UI overlays, and debug visualization.
line2DEML(v0, v1, vertex_color, depth) — 2D line with fvec2 endpoints and fvec4 color| Method | Description |
|---|---|
quad2DEML(quadrect, uvA, uvB, depth) | Basic screen quad with dual UV sets |
quad2DEML2(quadrect, uvA, uvB, depth) | Variant 2 |
quad2DEMLCCL(quadrect, uvA, uvB, depth) | CCL winding variant |
quad2D(quadrect, uvA, uvB, depth) | Auto-winding (uses RenderingConventions) |
fullscreenQuad(uvA, uvB, depth) | Fills entire viewport |
Parameters:
quadrect — fvec4(x, y, w, h) in screen spaceuvA, uvB — fvec4(u0, v0, u1, v1) UV rectangles for two texture channelsdepth — Z depth value (default 0.0)quad3DEML(V0, V1, V2, V3, Uv0, Uv1, Uv2, Uv3, color) — 4 fvec3 vertices with per-vertex UV and uint32 colorquad2DEML(v0..v3, uvA0..uvA3, uvB0..uvB3, color, depth) — 4 fvec2 vertices, 4+4 fvec2 UVs per channel, fvec4 colorquad2DEMLTiled(quadrect, uvA, uvB, numtiles) — subdivided quad (legacy)quadTiled2D(quadrect, uvA, uvB, numtiles) — auto-winding tiled quadquad2D, fullscreenQuad, quadTiled2D) adjust vertex order based on Context::renderingConventions() to handle Y-flip differences between backendsSVtxV12C4T16 vertex format internally (position + color + 2 UV sets)// Post-processing fullscreen pass
auto* DWI = ctx->DWI();
auto* FXI = ctx->FXI();
FXI->BeginBlock(technique, rcid);
DWI->fullscreenQuad(); // Renders a screen-filling quad
FXI->EndBlock();
// Debug overlay quad
DWI->quad2D(
fvec4(10, 10, 200, 200), // screen rect
fvec4(0, 0, 1, 1), // UV channel A
fvec4(0, 0, 1, 1), // UV channel B
0.0f // depth
);
dwi = ctx.DWI
# Fullscreen quad (most common for post-processing)
dwi.fullscreenQuad()
dwi.fullscreenQuad(fvec4(0,0,1,1), fvec4(0,0,1,1), 0.0)
# Screen-space quad
dwi.quad2D(fvec4(x, y, w, h))
dwi.quad2D(fvec4(x, y, w, h), fvec4(u0,v0,u1,v1), fvec4(u0,v0,u1,v1), depth)
dwi.h:9–56RenderingConventions in gfxenv.h:166–182SVtxV12C4T16 in gfxvtxbuf_structs.hpyext_gfx.cpp:298–322