# init, capture, skills

<!-- registry-items: allow=blank,landscape-4k,portrait-4k,square-4k,media-use -->

Scaffolding commands. Use these instead of creating files by hand — they set up the right file structure, copy media, run transcription, and install AI coding skills.

## init

```bash
npx hyperframes init my-video                                    # centered blank (TTY: wizard)
npx hyperframes init my-video --example warm-grain               # pick an example
npx hyperframes init my-video --resolution portrait
npx hyperframes init my-video --video clip.mp4                   # with video file
npx hyperframes init my-video --audio track.mp3                  # with audio file
npx hyperframes init my-video --tailwind                         # Tailwind v4 browser runtime
npx hyperframes init my-video --non-interactive                  # CI — flag-only, same blank
```

**Default depends on TTY**: in a terminal, the CLI prompts for example/options (default: centered blank). Outside a TTY (CI, agents, piped output) it auto-switches to non-interactive and scaffolds that blank. Pass `--example` only to start from a named example. Pass `--non-interactive` to force flag-only mode on a TTY.

Templates: `blank`, `warm-grain`, `play-mode`, `swiss-grid`, `vignelli`, `decision-tree`, `kinetic-type`, `product-promo`, `nyt-graph`. (The closed set of `hyperframes:example` items in `registry/registry.json` plus the bundled `blank` template. `hyperframes catalog` does not list examples — its `--type` takes only `block` or `component` — so this list has no live equivalent and is checked by `bun run lint:skills`.)

Other useful flags:

- `--resolution` — preset: `landscape` (1920×1080), `portrait` (1080×1920), `landscape-4k`, `portrait-4k`, `square` (1080×1080), `square-4k`. Aliases: `1080p`, `4k`, `uhd`, `1080p-square`, `4k-square`.
- `--skill=<slug>` — record the owning skill (e.g. `hyperframes`) in `hyperframes.json`, so every later render of this project — re-renders, `npm run render`, `--batch` — is attributed to it on anonymous telemetry without re-passing the flag. You rarely pass it by hand.
- `--skip-skills` — **temporarily ignored**: `init` always checks AI coding skills against GitHub while the skills.sh registry catches up. To opt out (CI/tests), set the `HYPERFRAMES_SKIP_SKILLS=1` env var instead.
- `--skip-transcribe` — don't auto-transcribe `--audio` / `--video` with Whisper.
- `--model`, `--language` — Whisper model / language for the auto-transcription.

**Tailwind** (`--tailwind`): the scaffold pins the Tailwind v4 browser runtime, not Studio's Tailwind v3 setup.

- Keep the pinned runtime; never swap in `cdn.tailwindcss.com`. Renders wait for `window.__tailwindReady` before frame 0. For offline or locked-down renders, compile to CSS and ship the stylesheet.
- Theme and custom utilities go in `<style type="text/tailwindcss">` with `@theme` and `@utility`. No v3 `@tailwind base/components/utilities`, no `tailwind.config.js`; `@config` and `@plugin` abort the browser compile.
- Write complete class tokens in the HTML. The runtime only sees classes it can scan, so never build a render-critical class name at seek time (`bg-${color}-500`); vary items through CSS variables (`style="--i: 2"`) or data variants.
- Tailwind is for static layout and style; timing stays in GSAP or another seekable runtime.

When `--audio` or `--video` is supplied, `init` transcribes the file with Whisper. For voice/model selection see the `media-use` skill.

## capture

```bash
npx hyperframes capture https://stripe.com                  # scaffold from a website
npx hyperframes capture https://linear.app -o linear-video  # custom output directory
npx hyperframes capture https://example.com --json          # JSON output for agents
npx hyperframes capture https://example.com --skip-assets   # skip image/SVG download
npx hyperframes capture https://example.com --skip-vision   # skip optional AI captions
npx hyperframes capture https://example.com --max-screenshots 12
npx hyperframes capture https://example.com --timeout 60000 # page-load timeout in ms
npx hyperframes capture https://example.com --capture-budget 90000 # post-navigation budget
```

Captures a live URL as an editable HyperFrames project: screenshots become layered scenes, assets are downloaded locally, and the result is a normal project you can `lint` / `preview` / `render`. Use this when the user supplies a URL as the starting point for a video.

`--timeout` bounds page navigation; `--capture-budget` is the separate cooperative budget for work
after navigation (fonts, assets, vision, and contact sheets). The latter is not a hard wall-clock
watchdog and cannot interrupt native work already in flight. An outer caller deadline is therefore a
third, distinct timeout. An outer caller timeout leaves the capture result unknown; it does not prove
HyperFrames hung or that the navigation timeout should be increased. Preserve the last phase and
classify the boundary that fired. `--skip-vision` disables only optional AI image captioning.

For agents, use `--json`. The result includes `ok`, warnings, and `lastPhase`. The command also emits
stable `HYPERFRAMES_CAPTURE_PHASE` records so a watchdog can report the last started, completed, or
degraded phase without retaining sensitive payloads.

Treat a non-zero exit, JSON `ok: false`, or an output `BLOCKED.md` as a **hard stop**. Do not render,
build, or infer brand/design data from partial files in a blocked capture. A successful capture may
degrade an optional phase within budget, but its structural output still has to be usable. Exit zero and file existence alone are not semantic success: require the current
invocation's JSON `ok: true`, no `BLOCKED.md`, and artifacts usable for the video. Run each retry
into a fresh output directory; never merge or reuse a blocked attempt's partial output.

## skills

```bash
npx hyperframes skills    # install HyperFrames skills for AI coding tools
```

One-time setup that adds the HyperFrames skill pack (/hyperframes with its references, /hyperframes-cli, /hyperframes-registry, /media-use, /figma, /slideshow, /remotion-to-hyperframes) to the local AI coding environment so agents follow the framework conventions. Re-run after major HyperFrames upgrades.