iwsdk-debug
Skill1 starsChanged 31 days agoMITView on GitHub
Diagnose and verify real-time IWSDK behavior with bounded ECS pause, query, snapshot, step, and diff loops. Use for physics, animation, collision, interaction, or state bugs that cannot be proven from a screenshot.
In this repository58 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- cf-cliSkill
- durable-objectsSkill
- game-gdd-authorSkill
- game-playtestSkill
- pmndrs-handleSkill
- pmndrs-pointer-eventsSkill
- pmndrs-uikitSkill
- prototype-fastSkill
- pwa-offlineSkill
- review-game-trailerSkill
- scenario-blender-expertSkill
- scenario-blender-retopologySkill
- scenario-blender-uv-bakingSkill
- threejs-perf-loadingSkill
- webapp-testingSkill
- workers-best-practicesSkill
- wranglerSkill
- xr-perceptual-constraints-displays-and-trackingSkill
- xr-platforms-audio-and-comfortSkill
- xr-referenceSkill
- xr-rendering-input-and-spatial-understandingSkill
- xr-ux-design-and-performance-budgetingSkill
- horizon-launch-roadmapSkill
- horizon-store-listing-assetsSkill
- horizon-web-store-publishSkill
- horizon-web-vrc-checklistSkill
- iwsdk-asset-pipelineSkill
- iwsdk-dev-workflowSkill
- iwsdk-performanceSkill
- iwsdk-project-setupSkill
- iwsdk-quest-device-testingSkill
- iwsdk-testingSkill
- iwsdk-vr-glassesSkill
- quest-browser-webxrSkill
- iwsdk-build-modelSkill
- iwsdk-compose-sceneSkill
- iwsdk-depth-occlusionSkill
- iwsdk-devSkill
- iwsdk-grabSkill
- iwsdk-hostingSkill
- iwsdk-migrate-0-5Skill
- iwsdk-native-xr-testSkill
- iwsdk-physicsSkill
- iwsdk-pwa-packagingSkill
- iwsdk-raySkill
- iwsdk-uiSkill
- hz-immersive-designerSkill
- hz-iwsdk-webxrSkill
- hz-new-project-creationSkill
- hz-perfetto-debugSkill
- hz-quest-verify-firstSkill
- hz-simpleperf-debugSkill
- hz-store-pwaSkill
- hz-store-submitSkill
- hz-vr-debugSkill
- metavr-cliSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Use the current request. The goal is a reproducible observation, the smallest supported fix, and one matching replay—not an open-ended …
- Distinctive: line 14
Start the managed app once and enter XR only if required.
- Very specific: line 38
Use the common authored physics forms directly instead of dumping scene capabilities: a detailed static environment normally pairs …
- Distinctive: line 47
Do not inspect IWSDK framework source, Vite plugin internals, or CLI help when a missing authored component already explains the measured …
- Distinctive: line 56
After editing, do not run an intermediate replay.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
IWSDK Runtime Debugging
Use the current request. The goal is a reproducible observation, the smallest supported fix, and one matching replay—not an open-ended trace session.
1. Establish one failing observation
Start the managed app once and enter XR only if required. Locate and query the named entities and relevant systems. Freeze time as early as possible:
ecs_pauseecs_snapshotwith labelbefore- trigger or prepare the failing transition
ecs_stepfor 1-3 frames at0.016ecs_snapshotwith labelafterecs_difffrombeforetoafter
Query the affected entity after the diff. For a startup failure that has already occurred, record the bad live transform first, then use the next clean restart for the frame-stepped observation. A screenshot can clarify geometry but is not a substitute for ECS evidence.
2. Form and test one hypothesis
Inspect only the configuration, scene node, component, or system implicated by the measured change. Check the invariant before editing: colliders need both body and shape, dynamic motion needs the correct state, interactions need the matching input/component path, and animations need the responsible system to be running. Do not browse framework internals while an authored invariant already explains the evidence.
Use the common authored physics forms directly instead of dumping scene
capabilities: a detailed static environment normally pairs
PhysicsBody: { "state": "STATIC" } with
PhysicsShape: { "shape": "TriMesh" }; dynamic objects use a supported convex
or primitive shape. Query live components after reload to verify registration.
Make the smallest project edit that restores that invariant. Avoid unrelated cleanup, visual redesign, or defensive rewrites.
Do not inspect IWSDK framework source, Vite plugin internals, or CLI help when a missing authored component already explains the measured state. Use public component data and the project files as the debugging boundary. For a local authored-component fix, do not request the full scene capability or scene-state payload; the source diff, production build, and focused live entity query provide the required evidence.
3. Replay the same probe
After editing, do not run an intermediate replay. Run the production build, restart once, and repeat the same pause → snapshot → step → snapshot → diff sequence with the same labels and timestep. Compare the same fields and query the final entity. This gives exactly two probes: one failing observation and one post-build replay. Capture one final runtime screenshot if position or collision is visually relevant.
The build is the production-build proof; do not launch a second Vite preview
server. Replay with the same managed IWSDK dev runtime. Save file evidence with
npx @iwsdk/cli browser screenshot --output-file <path> rather than discovering the
screenshot CLI after the debug loop. When ECS evidence already proves the fix,
do not open or read the PNG back into model context; leave it for evaluator or
human review. Inspect one compact image only when the spatial result cannot be
established from the measured state.
Make at most one focused correction. Stop when the original symptom no longer
reproduces and the measured state agrees with the intended behavior. Call
ecs_resume before finishing and leave input buttons released.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

