القائمة
Capture and analyze Imla Android renderer smoke traces with the repo-local Perfetto config, SQL queries, and Android Performance Analyzer workflow.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Capture Imla Android screenshots through the repo-local half-size ADB screenshot wrapper to reduce visual-inspection token cost while preserving enough detail for renderer checks.
Run Imla ADB commands through the repo-local timeout wrapper so physical-device installs, launches, input, screenshots, package-manager calls, and diagnostics fail fast instead of hanging.
Read Imla scene renderer counters from repo-local Perfetto analysis. Use when evaluating render reasons, root captures, slot content captures, mask/clip captures, geometry refreshes, scene renders, or whether an interaction is content-capture-heavy versus geometry-only.
Compare Imla Android renderer screenshots with repo-local metrics and visual diff artifacts. Use when visual verification needs screenshot comparison, noise-tolerant renderer diffs, blur/mask/clip regression checks, or threshold-gated image comparison.
Use the repo-local Kotlin outline tool to inspect Kotlin file or directory structure before broad source reads, especially in Imla renderer, Compose modifier, and Android UI code.
| name | imla_perfetto_feedback |
| description | Capture and analyze Imla Android renderer smoke traces with the repo-local Perfetto config, SQL queries, and Android Performance Analyzer workflow. |
Use this skill before and after renderer, blur, GL threading, capture, or Compose modifier changes that need runtime feedback beyond screenshots.
The repo-local tool is:
tools/imla-perfetto-feedback
It wraps the checked-in Perfetto smoke config, named SQL queries, screenshot capture, and optional Android Performance Analyzer opening. The extensionless shell wrapper is the stable entrypoint; the .py file is the implementation detail.
Analyze an existing trace:
tools/imla-perfetto-feedback analyze diagnostics/apa/traces/imla-smoke-2026-05-23.perfetto-trace
Capture a new smoke trace on the active emulator, exercise the demo, pull a screenshot, analyze it, and open it in APA:
tools/imla-perfetto-feedback smoke --device emulator-5580 --open-apa
Capture without scripted input:
tools/imla-perfetto-feedback capture --device emulator-5580 --no-exercise
Open a trace in Android Performance Analyzer:
tools/imla-perfetto-feedback open diagnostics/apa/traces/<trace>.perfetto-trace
Start investigation from these labels when present:
ImlaSceneCoordinator#commitRootCaptureImlaSceneCoordinator#refreshSlotGeometryRenderableRootLayer#captureLayerGraphicsLayerTexture#captureGraphicsLayerGraphicsLayerTexture#captureCanvasSceneLayerRepository#captureSlotContentSceneMaskRepository#captureSlotMasksSceneGlRenderer#renderRenderer2D#endSceneRenderer2D#uploadVerticesvboSetDataglBufferDataKey threads are the main app thread, Android RenderThread, and Imla GLUiLayerRender.
./gradlew -q :app:assembleDebug then install or launch via ADB.tools/imla-perfetto-feedback smoke --device <serial>.perfetto-trace-analysis and perfetto-sql for deeper AI-assisted trace questions; this repo tool provides the repeatable capture/query baseline.The scene counter summary is written to 60_scene_counters.txt when scene counters are present.