# Universal Visual Computing

> Discover and compose safe raster, vector, CV, OCR, video, procedural and optional model capabilities.

- Skill: `samraddhashrivastavatech/universal-visual-computing` (Agent Skill, multi-file: 24 files)
- Install (CLI): `npx skillmds@latest add samraddhashrivastavatech/universal-visual-computing`
- Raw SKILL.md: https://api.skillmd.com/api/skills/samraddhashrivastavatech/universal-visual-computing/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: SAMRADDHASHRIVASTAVATECH (https://skillmd.com/u/samraddhashrivastavatech)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/samraddhashrivastavatech/universal-visual-computing

---

# Universal Visual Computing Skill
1. Call capability discovery; never assume GPU, executable or model availability.
2. Import sources into the artifact store and pass IDs through DAGs.
3. Prefer the lightest deterministic operation. Combine nodes into a validated acyclic graph.
4. Never fabricate detections/OCR. Return provider, parameters, confidence and artifacts.
5. Preserve color, alpha, dimensions and provenance. Keep editable SVG/scene source when requested.
6. Route CPU by default; use GPU only if discovered and workload warrants it. Degrade to CPU or return an install plan.
7. Optional installs require approval, license/disk/RAM/VRAM disclosure and a rollback plan.
8. For screens use event/accessibility metadata and changed-region analysis before semantic models.
9. No arbitrary command execution, hidden downloads, kernel capture, injection or bypass.

## Domain authority
Use `research/UNIVERSAL_VISUAL_DISCIPLINES.md` to map professional deliverables to typed computation. Expose capability packs rather than pretending unavailable software, models, hardware, physical processes, or specialist validation succeeded.