使用方法
- 复制安装命令,在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物,并对照来源检查输出。
npx skills add https://github.com/nonaxanon/r3f-skills --skill r3f-debug-profiler开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况,不等于独立质量或安全测试。
按源仓库内容呈现,保留标题、案例、代码、表格、链接以及原文引用的演示图片。
r3f-debug-profiler · STUB
State
This skill is a stub. The seam is defined; the harness and checklists are not written yet. Do not present its output as a completed profiling phase. If you must measure now, read gl.info directly from a useFrame and report the raw numbers, and note in the director's risk ledger that this skill was unbuilt.
Intended purpose
Make scene problems measured, not guessed. Two halves:
- Correctness triage — a runbook for the blank/black canvas (first console error, render loop
running, canvas drawing-buffer size vs CSS size, camera framing/near-far, lights/material side, asset URLs/CORS) and for resize/mobile/DPR bugs.
- Performance profiling — a self-reporting diagnostics global populated from a
useFrame
hook (gl.info.render.calls/triangles, gl.info.memory.geometries/textures, frame time, canvas size/DPR) so an out-of-process script can read the scene's vitals, plus a headless canvas inspector that screenshots the <canvas>, proves it isn't blank by pixel variance, captures console/page errors, and waits for real frame progress before asserting.
Intended references / scripts (to author)
references/checklists/scene-debugging.md— the black-canvas runbook.references/checklists/performance-profile.md— the measurement recipe and what to capture.scripts/inspect-r3f-canvas.mjs— headless Chromium canvas-pixel + diagnostics-global reader.- A
useFramediagnostics-publisher snippet writingwindow.__R3F_DIAGNOSTICS__.
Dependency
Can run after any frame-budget-moving change (it is "measure now," not a single slot). Pairs with r3f-scene-architect (leak/draw-call findings) and r3f-graphics-builder (perf-safe detail).