Fit starts with the rendering problem
The useful question is not whether a job title appears on a list. The useful question is whether the work has a Three.js WebGPU rendering problem with an observable result, an owner, and a way to verify it. The pack is designed for work where a coding agent must reason about geometry, materials, motion, cameras, shared passes, output transforms, or validation rather than paste an isolated snippet.
The Choose Skills router starts from a workload profile and the earliest missing causal layer. It selects the smallest expert intersection, records tempting but unnecessary skills, and routes unsupported application concerns away from the graphics pack.
For graphics engineers
Choose the graphics-engineer path when the hard questions involve architecture and ownership. Typical work includes deciding whether compute is justified, identifying shared depth or velocity consumers, assigning one tone-map owner, comparing a minimal-forward graph with MRT, and defining target-specific evidence.
This path assumes that precise language about coordinate spaces, resource lifetime, invalidation, timing, and error is useful. It does not assume that every scene needs the most complex pipeline.
For technical artists
Choose the technical-artist path when the desired result must be decomposed into silhouette, material response, motion, camera, and final-image causes. The pack provides inspectable TSL material mechanisms, procedural construction patterns, diagnostic views, and fixed-capture contracts.
The goal is not to add effects until an image looks busy. It is to establish which layer owns the visible result, verify that layer, and only then add downstream consumers such as bloom, AO, or grading.
For Three.js developers moving to WebGPU
Choose the WebGPU adoption path when an existing Three.js project contains WebGLRenderer, GLSL ShaderMaterial, EffectComposer, or readback assumptions that need deliberate migration.
The path connects migration guides with current WebGPURenderer, TSL, node material, RenderPipeline, and validation contracts. It treats the installed Three.js package as the API gate and does not claim that changing an import completes the transition.
For AI coding-agent users
Choose the agent-user path when the main problem is getting a coding agent to route correctly, load minimal context, respect project ownership, and return inspectable evidence.
The pack gives an agent explicit constraints and runnable examples. It does not turn a model into an infallible graphics engineer. Project source, current dependencies, target devices, and verification results remain authoritative.
Choose by workload when the title is not enough
The same person may need different paths on different days. A technical artist designing a material system may use the material route, then switch to the graphics-engineer route when integrating shared depth and output ownership. A game developer may use the browser-game workflow for camera, effects, shadows, and quality tiers. A commerce team may use the product-configurator workflow to protect part identity, material response, and color output.
Start with the audience page that matches the immediate decision. Then use Choose Skills to select only the causal owners required by the scene.
What this page does not claim
These audience descriptions are inferred from the checked-in contracts, examples, and route boundaries. They are not customer segmentation research, testimonials, or adoption statistics. Final Image Flight has source-matched accepted evidence for its composed rendering contract. It cannot establish productivity, revenue, migration success, or performance on an unnamed target.
The pack also does not own generic application architecture, DOM UI, accessibility, live-data transport, asset-pipeline preparation, game logic, or commerce systems. The relevant audience page names those boundaries directly so that the coding agent does not quietly absorb unrelated responsibilities.