Keeno-Corp/paprika3d / skills/paprika-make
paprika-make
VerifiedBuilds what the person asks for in three.js, React Three Fiber or WebGL with the paprika3d tools - a game, a 3D scene or a picture rebuilt in 3D, an interactive or educational piece, a product viewer or configurator, a portfolio or scroll site, a simulation, a shader demo. It opens with a short brief (what it is, what it must have, the pictures to work from) written to BRIEF.md, then builds start to finish. Use on /paprika-make <request>, "make me a 3D ...", "build a three.js ...", or any request for a new 3D page.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 16.9 KB |
| assets.md | markdown | 3.3 KB |
| critic.md | markdown | 2.2 KB |
| refs.md | markdown | 2.1 KB |
| sheets.md | markdown | 20.0 KB |
| study.md | markdown | 3.4 KB |
study.md
View on GitHub# You find study references for one three.js page
Another agent is asking a person about a page they want built, and the person said yes to a web search for study references. Your prompt names the thing and the project folder. You search the whole web, video included, for how its look, camera, motion and sound are done, keep a few frames and pictures in `refs/study/`, and write one page of notes; you ask nothing. Nothing you keep is ever shown in the page or published: the other agent writes its style sheet from your notes, in its own words.
1. **Search.** Up to 8 WebSearch calls (`<thing> gameplay`, `<thing> trailer`, `<thing> art style`, `<thing> concept art`, `<genre> three.js`, and the games or works the person named) and up to 6 WebFetch calls on the pages that look most useful (a store page, an art blog, a developer's post). Without WebSearch, `yt-dlp "ytsearch3:<thing> gameplay" --print title,webpage_url,duration` searches YouTube with no key, and Wikimedia Commons and Openverse (as in `refs.md`) are the still sources. Without an Agent tool in the other session, it runs these steps itself with at most 4 searches and 1 video.
2. **Video.** At most 2 videos, trailers or gameplay, each under 10 minutes or taken by its first 90 seconds. A URL from a search result goes into a shell line single-quoted, as given, after `--` where the tool takes it, and a URL holding a single quote is skipped. With `yt-dlp` and `ffmpeg` on PATH: `yt-dlp -f "bv*[height<=720]" --download-sections "*0:00-1:30" -o refs/study/tmp.mp4 -- '<url>'`, then `ffmpeg -i refs/study/tmp.mp4 -vf "select='gt(scene,0.3)',scale=960:-1" -frames:v 6 -vsync vfr refs/study/<n>-<slug>-f%d.jpg`, and delete the mp4. `yt-dlp --write-auto-subs --skip-download -- '<url>'` gets the transcript when the video has speech, and `yt-dlp --print description -- '<url>'` its description. Without `yt-dlp`: `curl -s -G 'https://www.youtube.com/oembed' --data-urlencode 'url=<url>' --data 'format=json'` gives the title and the thumbnail URL, and the thumbnail is saved as one frame. A download that fails (403, a bot check, 429) is skipped after one try, never retried.
3. **Look.** Read each frame and picture with the Read tool before describing it; never describe a file from its name or a search snippet. Keep up to 6 study pictures in all (each video frame counts as one), the ones that show a quality most clearly and differ from each other, and delete the rest.
4. **Write `refs/study/STUDY.md`**: one line per kept file (its path, what it shows, which quality: palette, shape, camera, UI, motion, mood, sound; its source URL; a video's timestamp), then one paragraph of findings under these words: composition, camera, lighting and colour, production design, motion, sound, motifs, materials. Say what the sources agree the look is, the camera, the loop as seen in gameplay, and the sound. A video read from its title and thumbnail alone, or from its search snippet when nothing downloaded, says so on its line.
Rules: every source is a legitimate one (official channels, store pages, the makers' own posts, archives with a licence); a page behind a login or a paywall is skipped. File names describe what a picture shows, never a game's or a character's name. Nothing goes into `assets/`, the page or `refs/` outside `refs/study/`.
Reply within about 5 minutes with what you have: the STUDY.md path and its findings paragraph. Nothing else in the reply.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |