Guide · Agent skill
How to use the 3D Wood Lighting Scorecard agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The 3D Wood Lighting Scorecard is an agent skill that tells an AI agent how to light carved wood in a dark room using three.js, a web 3D library. It also gives the agent a headless browser scorecard that measures the render and scores it out of 10 against a written rubric. This guide explains what the skill does, how to install it, how to ask for it, and what it needs before you run it.
3D Wood Lighting ScorecardMengTo/Skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the 3d-wood-lighting-scorecard ZIP from this skill's page on the site.
- 2Unzip into the skills folder. Unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
- 3Check the requirements. Confirm Node.js, playwright-core or playwright, a Chromium browser, and network access for three.js r169 are available.
- 4Start the local server. Serve demo.html on http://localhost:8099, since the scorecard script expects that address.
- 5Ask for a lit wood scene. Describe the wood, the window light and the target score, for example getting the lighting to 8.
- 6Review the scores. Check the per-item scores from the scorecard and the saved references/last-scorecard.json file.
What the skill makes the agent do
Once installed, the agent can build a lit 3D wood scene in three.js with procedural grain, one window acting as the main light, soft shadows, light shafts and dust. It then renders the scene in a headless browser, measures the image with a fixed set of metrics, scores it out of 10 against a written rubric, and adjusts the lighting until each item reaches a target score.
It is a specialised lighting and material workflow for one lit interior, not a general 3D tool. The output is code for the three.js stack.
Install the skill from the ZIP
Download the ZIP from the skill's page on this site and unzip it. For a personal install, unzip it into ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Single project: .claude/skills/ inside the project
- Claude apps: Settings → Capabilities → Skills
Check the requirements before you run it
The skill needs Node.js, the npm packages playwright-core or playwright, and a Chromium browser (Chrome for Testing works). It also needs a local web server that serves demo.html on http://localhost:8099, because the scorecard script expects that address.
The scene loads three.js r169 from unpkg, so the machine needs network access for the library. The skill runs scripts and uses the network, so read the safety notes below first.
- Node.js
- npm packages: playwright-core or playwright
- A Chromium browser (Chrome for Testing works)
- A local web server for demo.html at http://localhost:8099
- Network access to load three.js r169 from unpkg
Trigger the skill with a natural request
Describe the wood and the look you want, and ask for a score. The skill's description names these situations: oiled timber, kumiko, joinery and workshop scenes; wood that reads flat, plastic or washed out; volumetrics that lift the blacks; and requests such as "score this out of 10" or "get the lighting to 8".
You can also name the skill directly with $3d-wood-lighting-scorecard, as the project's prompt file does.
- Light this carved timber like a workshop bench, one window key, and score it out of 10.
- This oiled wood looks plastic. Fix the grain under raking light and get the lighting to 8.
- Add faint light shafts and dust to this joinery scene without washing out the blacks.
Use $3d-wood-lighting-scorecard to light this carved timber like a Kibori station: one warm window key in a dark room, grain that shows under raking light, soft shadows that touch, faint shafts with dust, and score the render out of 10 until every item reaches 8.What to expect from the output
The agent follows a build order: it sets up the renderer, places one window that the key light and shafts come from, adds the key, fill, rim and hemisphere lights, builds the procedural wood, then adds shafts, dust and a final tone-mapping pass. It then runs references/scorecard.mjs to measure the frame.
The rubric has eight items, each scored 0 to 10 from measured sub-metrics. The agent keeps adjusting until each item reaches the target you set, such as 8.
- Scene code built in three.js, with procedural grain rather than image textures
- Measured scores per rubric item from references/scorecard.mjs
- A saved reference/last-scorecard.json file with the most recent results
Safety notes: scripts and network
The skill runs scripts. references/scorecard.mjs launches a headless browser and executes the bundled scripts, so run it only on machines and pages you trust. Do not run it on untrusted pages.
The scene loads three.js from unpkg over the network. Plan for that if your machine is offline or behind a strict firewall.
- Do not run scorecard.mjs on untrusted pages or machines.
- The skill launches a headless browser and uses the network to load three.js.
Tips for better results
Set the target score up front. Asking for 8 gives the agent a clear stopping point.
Pick the right wood type. The skill is written for procedural wood. If you need scanned or painted PBR texture maps, the skill points to a different skill, 3d-high-resolution-textures, for that work.
Use it for the lighting layer. For sun shafts cut by real trees and rooflines, the skill points to 3d-sky-rays. For a bright rim on dark metal, it points to 3d-specular-rim-glow.
- State a target score such as 8.
- Use procedural wood, not scanned textures.
- Pick 3d-high-resolution-textures for scanned or painted PBR maps.
- Pick 3d-sky-rays for sun shafts cut by real geometry.
What people use it for
- Lighting oiled hinoki or other carved timber on a workshop bench under one window.
- Making kumiko or joinery in a web 3D scene read as wood rather than plastic.
- Adding faint light shafts and dust to a dark interior without lifting the blacks.
- Getting repeatable scores for a render so changes can be compared numerically.
Frequently asked questions
How do I install the 3D Wood Lighting Scorecard skill?
Download the ZIP from this skill's page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps, you can upload it under Settings → Capabilities → Skills instead.
What do I need to run the 3D Wood Lighting Scorecard?
You need Node.js, the npm packages playwright-core or playwright, and a Chromium browser such as Chrome for Testing. You also need a local web server serving demo.html on http://localhost:8099 and network access to load three.js r169 from unpkg.
Is the 3D Wood Lighting Scorecard free?
The skill is published under the MIT licence. This guide does not state any price, so check the item's page on this site for current terms.
Does the skill work without three.js?
No. The output is code for three.js and WebGL, so it is not useful for projects that do not run that stack.
Can I use scanned or painted textures with this skill?
Not with this one. It uses procedural wood. For scanned or painted PBR maps, the skill points to 3d-high-resolution-textures.
Is it safe to run scorecard.mjs?
Only on machines and pages you trust. The script launches a headless browser and executes the bundled scripts, so it should not be run on untrusted pages.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you build web 3D scenes in three.js and want realistic wood, timber or joinery lighting that you can measure and iterate on with numeric targets. Pick something else if you do not run three.js or WebGL, need scanned or painted PBR texture pipelines, or do not want an agent running scripts and a headless browser on your machine.
Download the skill