home / skills / plurigrid / asi / ar
This skill helps you design and implement web and native augmented reality experiences with stable world-locked content, proper permissions, and safe
npx playbooks add skill plurigrid/asi --skill arReview the files below or copy the command above to add this skill to your agents.
---
name: ar
description: Augmented reality (AR) reality tech. Passthrough overlays, hit-test/anchors, occlusion, lighting, camera permissions, and safety.
license: Apache-2.0
metadata:
trit: 0
version: "1.0.0"
bundle: reality-tech
---
# AR (Augmented Reality)
Use when the user is building AR on WebXR (immersive-ar) or native (ARKit/ARCore/OpenXR).
## Fast Defaults
- Prefer stable world-locked content; avoid screen-locked UI unless necessary
- Use clear permission prompts and visible camera-on indicators
- Start with: hit-test + placement + anchors + simple interaction
## AR Capability Checklist
- Plane / mesh detection
- Hit-test (placing content onto surfaces)
- Anchors (persistent placement)
- Occlusion (depth) and lighting estimation if available
- Safety boundaries (keep critical UI in view, avoid distraction)
## Debug Checklist
- Confirm session type (WebXR `immersive-ar`) or native runtime
- Verify tracking origin and world scale (1 unit = 1 meter)
- Confirm anchor lifetimes and relocalization handling
See also: `ar-vr-xr` for the full workflow and shared performance/comfort guidance.
This skill covers building practical augmented reality (AR) experiences for WebXR (immersive-ar) and native platforms (ARKit/ARCore/OpenXR). It focuses on core AR features like passthrough overlays, hit-testing, anchors, occlusion, lighting estimation, camera permissions, and safety practices. The guidance helps you start with stable world-locked content and grow to persistent, well-behaved AR scenes.
The skill inspects and guides implementation of surface detection (planes or meshes), hit-test placement, and anchors for persistent content. It also verifies occlusion/depth handling and lighting estimation when available, and enforces camera permission prompts and visible camera-on indicators. Debug checks validate session type, tracking origin, scale assumptions, and anchor relocalization behavior.
How do I choose between world-locked and screen-locked UI?
Prefer world-locked UI for spatial tasks; use screen-locked UI only for constant controls or when safety requires it.
What to test for anchors during relocalization?
Verify anchor lifetime, whether anchors survive app restarts or device relocalization, and fallback behavior if relocalization fails.