AgentStack
SKILL verified MIT Self-run

De7 Webgl Fallbacks

skill-rakibulism-agent-skills-os-de7-webgl-fallbacks · by rakibulism

Track 7 rail — WebGL fallbacks and adaptive quality. Device capability detection and tiering (high/medium/low/static), the quality knobs ranked by cost-effectiveness (pixel ratio first), runtime FPS-monitored degradation, designed static-tier and loading-state choreography, prefers-reduced-motion handling, and WebGL context-loss recovery. Use whenever a 3D/animated experience must work on low-end…

No reviews yet
0 installs
5 views
0.0% view→install

Install

$ agentstack add skill-rakibulism-agent-skills-os-de7-webgl-fallbacks

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of De7 Webgl Fallbacks? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

WebGL Fallbacks & Adaptive Quality

Tier the experience — same design intent at every tier. The low tier is a designed artifact, never an apology page.

1. Capability Detection at Boot

Classify high / medium / low / static from:

  • WebGL2 support; MAX_TEXTURE_SIZE; unmasked renderer string (WEBGL_debug_renderer_info) for known-weak GPU patterns
  • navigator.hardwareConcurrency, navigator.deviceMemory, connection type
  • prefers-reduced-motion → static tier (or opacity-only transitions) automatically

2. Quality Knobs — ranked by visual cost-effectiveness

  1. Renderer pixel ratio — the single biggest lever: renderer.setPixelRatio(Math.min(devicePixelRatio, tier === 'high' ? 2 : 1)). Dropping 2→1 quarters the fragment work; most users can't tell on motion-heavy scenes.
  2. Postprocessing off (bloom/DOF/SSAO are the most expensive per-pixel items).
  3. Particle counts ÷ 4; raymarch iterations ÷ 2 (de3-glsl-shaders); shadow maps → baked (de3-asset-pipeline) or off.
  4. Smaller texture set / mip bias; simpler materials (standard → lambert) as a last resort.

3. Runtime Adaptive Degradation

Don't trust the boot classification alone — monitor and react:

// rolling FPS average in the render loop
if (avgFPS  58 for 10s) cautiouslyStepUp();

drei's `` implements exactly this for R3F. Never let users watch a slideshow out of pride — degrade live.

4. The Static Tier — designed, not apologetic

A high-quality rendered still or short video of the scene, same layout, same copy — indistinguishable in screenshots from the live version. This is also the poster in the LCP strategy (de7-core-web-vitals), so it exists anyway. prefers-reduced-motion users land here by default.

5. Loading States Are Choreography

  • Branded progress tied to real asset loading (THREE.LoadingManager / drei useProgress) — never an indefinite spinner.
  • Reveal with a designed transition (fade/wipe from poster to canvas), dimensions matched exactly (CLS — de7-core-web-vitals).
  • Grace period: if load exceeds ~4s on the detected connection, stay on the static tier and offer the full experience as an opt-in.

6. Context Loss Is Not Optional

Mobile GPUs evict contexts routinely (backgrounding, memory pressure):

canvas.addEventListener('webglcontextlost', e => { e.preventDefault(); pauseLoop(); });
canvas.addEventListener('webglcontextrestored', () => { reinitScene(); resumeLoop(); });

Report loss events with GPU info to telemetry (de8-telemetry) — a device class losing contexts constantly means the tiering misclassifies it.

Ship Test

Actually run the low tier and reduced-motion mode before launch — on a real cheap Android, not a resized desktop window.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.