ztx-3d-point-cloud

Depth-aware point clouds. Invoke for precise implementation, responsive states, tests and integration into existing code; part of the Perfect_AI OMEGA no-MCP Codex skill collection.

SKILL.md

GitHub

Depth-aware point clouds

When to invoke

Use for tasks involving depth-aware point clouds, especially when user requests production-grade Persian/English UI, complex motion, 3D, mobile design, color selection or code review. This skill is authored locally, not third-party code.

Mandatory implementation procedure

  1. Inspect current routes/HTML/CSS/TS, component ownership, dependencies and existing tests. Preserve working behavior; write an inventory of affected files.
  2. Read ../perfect-ai-master/references/46-advanced-webgl-and-shader-systems.md in this installed pack. For complex projects also read 22-responsive-engineering-foundations.md, 36-advanced-color-science.md and 50-skill-composition-knowledge-routing.md as relevant, not all 50 modules blindly.
  3. Translate requirements into explicit semantic DOM, state machine, responsive layout, animation stages, or GPU scene data. Identify what works without JavaScript or WebGL.
  4. Implement real working code, minimal dependencies, error handling and cleanup. Choose exactly one motion owner for each CSS/Three property; ensure RTL direction and accessible actions.
  5. Define motion/input/readability fallback for 320px narrow, 400% zoom, no hover, reduced motion, failed network/GPU and short landscape viewport. Test scenarios that materially apply.
  6. Run syntax/typecheck/build and real browser tests if available. Distinguish passing tests from things only specified. Do not claim proprietary reference website source was extracted.

Concrete engineering rules

  • Layout: intrinsic CSS Grid/Flex, minmax(0,1fr), min-width:0, logical properties; named container queries when child allocated size controls layout.
  • Type and color: semantic tokens; clamp() without fixed-zoom lock; choose OKLCH scientifically and audit WCAG contrast; user-requested monochrome outranks auto colors.
  • Motion: deterministic progress, reversibility, cleanup, reduced-motion static state and no accidental transform conflicts. Camera/geometry truly 3D if claimed.
  • Interactive elements: visible focus, keyboard activation, touch target, loading/error/empty outcome; no placeholder CTA with no action.
  • Performance: dynamic imports for heavy runtimes, image aspect ratios, no offscreen infinite rAF, GPU quality tier and resource disposal.

Output contract

Produce implementation files with comments at non-obvious boundaries; tests for target capability; concise report of changed files, responsive matrix exercised, actual diagnostics and known limitations. Perfect_AI default is intentional monochrome.

References

  • Local master: skills/perfect-ai-master/references/46-advanced-webgl-and-shader-systems.md.
  • VibeFarsi local catalog (279 original specs retained): skills/perfect-ai-master/registry/ (fetch upstream via CLI only with user's permission, no MCP).
  • Reference APIs are linked in master guide; verify versions, avoid blind copying code from external sites.