Skip to content

Adonis0123/adonis-skills / skills/prompt-studio

prompt-studio

Flagged

Turn a one-line image or video request into a finished asset: route it to an image, a code-rendered Code Video or a Model Video, search many GitHub prompt libraries for proven image prompts, video-model prompts (Seedance, Kling, Veo, Hailuo/MiniMax, Wan, Grok Imagine) and code-video templates, show the best 5 with their samples, model, author and license, write a generation brief, hand it to an available channel (host tool, Codex/Grok CLI, API, Hyperframes), then check and log the result. Use whenever the user wants a cover, illustration, poster, thumbnail, infographic, animation, motion clip, AI video, or a better prompt for one, even if they only say 'make me an image'. Also for Chinese-language requests for WeChat covers, illustrations, still-to-motion clips, and image/video prompt optimization. Not for general chatbot or coding prompts, photo editing without generation, UI implementation, or slide decks.

references/code-video.md

View on GitHub
# Code Video

An LLM writes the composition; an engine renders frames. Every word and number is exact, and the cost is tokens plus local render time.

## Still to motion (动效 default)

When the user says 动效, 动起来, or "motionize this still" without asking for realistic motion or naming a video model, build a Code Video: it is free and every word and number stays exact. Rebuild the still's layout in code (or place the still as a background layer), animate transform and opacity only, keep the same palette and Style Pack, and default to a 3–5 s loop at the still's aspect ratio. Offer Model Video only as an upgrade, with a cost quote.

## Pick the engine

| Engine                                                         | Best for                                                                 | Needs                   | Notes                                                                                                                           |
| -------------------------------------------------------------- | ------------------------------------------------------------------------ | ----------------------- | ------------------------------------------------------------------------------------------------------------------------------- |
| Hyperframes (default)                                          | Launch clips, explainers, data motion, vertical shorts, captions, slides | Node.js 22+, FFmpeg     | ~450 registry blocks, built-in checks. Skill: `heygen-com/hyperframes`                                                          |
| Remotion                                                       | Work inside an existing React project, batch or cloud rendering          | Node.js, `create-video` | All `remotion` and `@remotion/*` packages on one exact version. Company use may need a paid license: check remotion.dev/license |
| p5.js + p5.brush (for example a local painted-animation skill) | Hand-painted watercolor cartoons, character shorts                       | Node, FFmpeg, Chrome    | Slow without a GPU                                                                                                              |
| Kinetic typography (for example a local kinetic-reel skill)    | Big-type portfolio reels, product promos                                 | Node, FFmpeg, three.js  | Motion on the beat                                                                                                              |

Hyperframes is the default. Use another engine only when its skill is already installed and the request fits its row (a painted short, a kinetic-type reel, an existing Remotion project). When the Hyperframes skill is not installed, read its text through `delivery.md` layer 2 and offer to install it afterwards.

A template is a starting structure, not a spec. When it asks for real screenshots, music, voice-over or a logo the user did not give, use clean placeholder UI (simple shapes, the product name as a wordmark), no audio, and say so in one line with the result. Ask before starting only when the request depends on an asset (their real app screens, their logo).

## Code-built still

For a poster or card whose text must be exact: write one HTML page at the final size (`width`/`height` in px on `body`), with the generated background as a CSS `background-image` and the text as real HTML with a bundled CJK font (`@font-face`). Render it with `npx hyperframes snapshot` on a one-frame composition, or open it in a browser and take a full-page screenshot at the exact size. Check with `studio.mjs check-image <file> --size WxH`.

## Setup

- `npx hyperframes init <absolute dir> --skip-skills` (or `HYPERFRAMES_SKIP_SKILLS=1`): without it, `init` installs Hyperframes' agent skills, which skips the install confirmation in `delivery.md` layer 3. Pass absolute paths to `init` and run every later `hyperframes` command with the project as its working directory (`npm --prefix <dir>` or `cd <dir> && ...` in the same command): a shell `cd` in a separate call may not persist, and a relative `init` then creates the project inside whatever repository the shell is in.
- No local Chrome: set `HYPERFRAMES_BROWSER_PATH` to an existing Chrome or Chromium; otherwise `check` downloads a headless shell and can time out.
- Chinese or other non-Latin text: bundle a font that covers it in `assets/fonts/` with `@font-face`; the built-in fonts are Latin only and render CJK punctuation as boxes.
- Open the preview before rendering and look at it once.
- `check` and `snapshot` may leave a headless browser running for a few minutes; before you finish, stop any process you started, by pid.

## Templates

Start with `studio.mjs recommend "<motion description>" --kind code-video`: it searches the Hyperframes catalog (posters and preview clips) together with code videos other people made with an agent. A Hyperframes result's name is what `add` takes. The CLI search below is the fallback and lists blocks the index may not have yet.

```bash
npx hyperframes catalog --query "<English description>" --json   # search, offline index
npx hyperframes add <name> --json                                   # copies editable source into compositions/
```

Before the first write, ask once: the project directory `init` creates, the CLI version (`npx hyperframes --version`), each block to `add` with the files it writes, and a one-line summary of any changed skill text (`delivery.md`). Do not run `init` before that yes. If `catalog` returns nothing (it can come back empty behind a proxy), write the composition from the Request Brief directly; do not loop on the search. Remotion has no add command: start from `npx create-video@latest --yes --blank <dir>` (fails inside an existing Git repo; pick a fresh directory).

## Request Brief fields (from the Hyperframes spec)

- GOAL: one sentence — what the viewer feels or does afterwards.
- OUTPUT: W×H and fps (default 1920×1080 at 30 fps; use the Style Pack preset when given).
- TIME: seconds × fps = frames; BPM and frames per beat if there is music.
- LOOK: background hex, ink colour, one accent and where it may appear, font + weight + px size, texture, grain or vignette percentages.
- OBJECTS: each thing on screen with recognisable material and detail.
- MOTION: named easings, spring damping, durations and staggers in frames.
- SHOTS: one line each — frame range, the single main action, what the viewer must read and for how long, sound, hand-off to the next shot.
- BANNED and QA: what must not appear; which frames to check and what each must show.

Facts on screen must come from the user or a cited source; mark illustrative numbers as schematic. Labels are facts too: when the user gives numbers without saying what they measure (Q1 120 万 …), label them neutrally (季度数据(万)) and offer the likely meaning (营收?) as a question in the same message, instead of writing 营收 into the film. Check three things separately before calling it done: the input numbers, every derived number (growth %, differences), and that the title and labels say only what the user said.

## Rules that keep quality high

- Every frame is a pure function of time: no `Date.now()`, unseeded `Math.random()`, network calls, timers, or infinite repeats.
- Remotion: animate with `useCurrentFrame()` + `interpolate()`; CSS transitions and animations do not render.
- Hyperframes: every timed element has `class="clip"`, `data-start`, `data-duration`; register one paused timeline per composition (`window.__timelines[id] = tl`); never tween `display`, `visibility` or `autoAlpha` on a `.clip` itself, animate its children (this overrides GSAP's general advice to prefer `autoAlpha`).
- GSAP: sequence with a timeline and position parameters, not chained delays; animate transform and opacity, not width, height, top or left.
- One main action per shot; the subject fills at least 40% of the frame; something new every 2–4 seconds.
- Text colour needs 4.5:1 contrast with its background; a brand accent that fails goes on shapes and highlights, not on text.
- Readable text: at least 28 px at 1080p, on screen for 0.5 s + 0.3 s per word. Too many words → cut words, never shrink type.
- Default AI look is banned unless asked for: a black or dark-navy void, glow, bloom, neon, purple-blue gradients, glass cards, everything fading in at once, three cards in a row, KPI grids, a logo parked in a corner.
- Narrative clips longer than ~15 s get music and sound effects by default. You cannot hear audio: measure it (`ffmpeg -i out.mp4 -af volumedetect -f null -`) and say it was measured, not listened to.

## Quality gate

1. Hyperframes: `npx hyperframes lint`, then `npx hyperframes check --json` must be clean (it reports unread lines, dead stretches, text below size, an empty first frame, silent films).
2. Look at frames: `npx hyperframes snapshot --review` (includes a 360 px phone width) or, for any engine, `ffmpeg -i out.mp4 -vf fps=1 frames/%03d.png` and open them.
3. Write findings as P0/P1/P2; fix every P0 before calling it done. Do not score the film out of 10.
4. Render: `npx hyperframes render --quality draft` while iterating, `--quality delivery --output out.mp4` at the end; confirm duration with `ffprobe`.
canvas-designanthropics/skillsCreate beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
docxanthropics/skillsUse this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pdfanthropics/skillsUse this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pptxanthropics/skillsUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates (.potx), layouts, speaker notes, or comments. Trigger whenever the user mentions "deck," "slides," "presentation," or references a .pptx or .potx filename, regardless of what they plan to do with the content afterward. If a .pptx or .potx file needs to be opened, created, or touched, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall