[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77WNK2WNJD98GEF1WW4CQ7M":3,"skill-same:01M77WNK2WNJD98GEF1WW4CQ7M":155},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":30,"ownerDescription":6,"updatedAt":31,"install":32,"resources":45},"01M77WNK2WNJD98GEF1WW4CQ7M","talking-head-recut",null,"Package an existing talking-head \u002F interview \u002F podcast video with timed, designed GRAPHIC OVERLAY cards — kinetic titles, lower-thirds, data callouts, quotes, side panels, picture-in-picture — synced to the transcript, on a 16:9 \u002F 9:16 \u002F 4:5 canvas of your choice; the clip plays untouched underneath. Trigger on \"graphic overlays\", \"on-screen graphics\", \"package \u002F dress up my video\". Not plain subtitles (\u002Fembedded-captions). Unclear → \u002Fhyperframes.","Reihantt6\u002Fsns-myagent","Reihantt6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F219822095?v=4","01M77KK94PV8B7C2D9XWXNS8DJ","Mohamad Reihan tegar prasetya","01M77KK958QFMTHYXGXFV7KZ79",2,"MIT","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\ndescription: \"Package an existing talking-head \u002F interview \u002F podcast video with timed, designed GRAPHIC OVERLAY cards — kinetic titles, lower-thirds, data callouts, quotes, side panels, picture-in-picture — synced to the transcript, on a 16:9 \u002F 9:16 \u002F 4:5 canvas of your choice; the clip plays untouched underneath. Trigger on \\\"graphic overlays\\\", \\\"on-screen graphics\\\", \\\"package \u002F dress up my video\\\". Not plain subtitles (\u002Fembedded-captions). Unclear → \u002Fhyperframes.\"\n---\n> **First, keep this skill fresh — run silently, don't ask:** `npx hyperframes skills update talking-head-recut`. A fast no-op when everything is current; otherwise it refreshes this skill plus the core domain skills it depends on before you rely on them.\n\n# Talking Head Recut\n\nTalking Head Recut takes a local video that **plays in full** and layers a sequence of\ntimed, designed **graphic cards** onto it — titles, lower-thirds, data callouts,\nquotes, side panels, picture-in-picture — synced to what's being said. The agent\ndesigns the cards (timing + content) and **writes each card's HTML directly in the\nconversation**, then assembles a single composition HTML and renders it to MP4 via\n`hyperframes`. There is no fixed archetype list and no prescribed card structure —\nthe overlays emerge from what the transcript actually says.\n\n> **The front door is `\u002Fhyperframes`.** This skill packages an **existing talking-head clip** with **designed graphic cards** (titles, lower-thirds, data callouts, quotes, side panels, PiP) — not plain captions (the spoken words as text). **The clip plays untouched.** Any other intent — plain subtitles, a standalone graphic, a from-scratch video — or any uncertainty → read `\u002Fhyperframes` first: the intent layer owns every route decision.\n\n> **Graphic-packaging sibling of `embedded-captions`.** Captions add the _spoken words_\n> as a readable subtitle; this adds _designed graphics_ on top of the playing video.\n> Plain subtitles → `embedded-captions`. Build a video from scratch → the creation\n> workflows (`product-launch-video` \u002F `faceless-explainer` \u002F …).\n\nRouted through `\u002Fhyperframes`, the intent layer confirms only the input (which clip) and **announces** the render-strategy questions as deferred asks — aspect, layout, style group, and card count stay at Step 7, where the probed footage and transcript ground the recommendations; the layer's run-shape questions don't apply. A `BRIEF.md`, when present, carries the confirmed input and any user notes — read it first.\n\nInspectable intermediate files in the work directory:\n\n- `metadata.json` — duration \u002F width \u002F height \u002F fps\n- `audio.mp3` — extracted audio\n- `transcript.json` — a flat **word array** `[{ text, start, end }, …]` (Whisper; no `segments`, no `words` wrapper)\n- `storyboard.json` — lightweight card outline (the agent's plan)\n- `public\u002Fcards\u002Fcard-XX.html` — one HTML fragment per card\n- `public\u002Findex.html` — final assembled composition\n- `output.mp4` — rendered video\n\n## CLI Resolution\n\n```bash\n# hyperframes — transcription (local Whisper) + rendering the assembled HTML to MP4\nnpx hyperframes --help\n```\n\nThis skill runs entirely on the **hyperframes** CLI plus system `ffmpeg` \u002F `ffprobe`.\nTranscription is local **Whisper** via `hyperframes transcribe` — no third-party\nservice, API key, or rate-limited proxy.\n\n## Workflow\n\n### 1. Check Environment\n\n```bash\nnpx hyperframes doctor          # ffmpeg, headless browser, render deps\n# confirm bundled assets:\nls \"\u003CSKILL_DIR>\u002Fassets\u002Ffonts\" \"\u003CSKILL_DIR>\u002Fassets\u002Fvendor\u002Fgsap.min.js\"\n```\n\nRequired:\n\n- `ffmpeg` \u002F `ffprobe` (system)\n- `\u003CSKILL_DIR>\u002Fassets\u002Ffonts\u002F*.woff2`, `\u003CSKILL_DIR>\u002Fassets\u002Fvendor\u002Fgsap.min.js` (bundled inside this skill, staged to work dir in Step 9)\n\nTranscription needs no key — `hyperframes transcribe` runs Whisper locally (Step 4).\n\nStrongly recommended on macOS for `hyperframes render`:\n\n```bash\nexport PRODUCER_BROWSER_GPU_MODE=hardware\n```\n\n### 2. Create a Work Directory\n\nAll artifacts live under `videos\u002F\u003Cproject-name>\u002F` — the same convention as the other\nvideo workflows (`product-launch-video` \u002F `faceless-explainer` \u002F `pr-to-video`). Keep\nthe cwd at the workspace root; everything below writes under this one subdirectory.\n\n```bash\nVIDEO_PATH=\"\u002Fabsolute\u002Fpath\u002Finput.mp4\"\nWORK_DIR=\"videos\u002F$(basename \"$VIDEO_PATH\" | sed 's\u002F\\.[^.]*$\u002F\u002F')\"\nmkdir -p \"$WORK_DIR\"\n```\n\n### 3. Extract Audio and Metadata\n\n```bash\n# metadata — duration \u002F width \u002F height \u002F fps\nffprobe -v error -select_streams v:0 \\\n  -show_entries stream=width,height,r_frame_rate \\\n  -show_entries format=duration -of json \"$VIDEO_PATH\" > \"$WORK_DIR\u002Fmetadata.json\"\n# audio\nffmpeg -y -i \"$VIDEO_PATH\" -vn -acodec libmp3lame -q:a 2 \"$WORK_DIR\u002Faudio.mp3\"\n```\n\nOutputs: `metadata.json` (read `width`\u002F`height`\u002F`duration`; fps = the `r_frame_rate`\nfraction evaluated, e.g. `30000\u002F1001 → 29.97`) + `audio.mp3`.\n\n### 4. Transcribe\n\n```bash\nnpx hyperframes transcribe \"$WORK_DIR\u002Faudio.mp3\" -d \"$WORK_DIR\" --json --model small.en\n```\n\nLocal **Whisper** — no API key, no proxy, no rate limit. Writes a word-level\n`transcript.json` into the work dir (word `text` + `start` \u002F `end` timestamps).\nRead it for the word \u002F sentence timings that drive card timing in Step 6; group\nwords into sentences yourself at punctuation \u002F pauses if you need segment-level\nchunks.\n\n**Clamp to media duration.** Whisper can return the final word's `end` a hair past the\nactual clip length — clamp every card `endSec` and `composition.durationSeconds` to the\n`metadata.json` duration, or the render will show a black tail past the video.\n\n### 5. Correct Transcript\n\n`transcript.json` is a **flat array of word objects** — `[{ \"text\": \"...\", \"start\": s, \"end\": s }, …]` (no `segments` array, no `words` wrapper; the per-word key is **`text`**). Read it and fix obvious ASR errors:\n\n- Homophones, product names, technical terms, punctuation\n- Edit a word's `text` in place; **preserve its `start` \u002F `end`** timestamps\n- There is no pre-grouped `segments` array — **group words into sentences yourself** (split at terminal punctuation \u002F pauses) when you need segment-level chunks for card timing\n\n### 6. Draft a Lightweight Storyboard (in chat)\n\n**No CLI involved.** Read `transcript.json` + `metadata.json` and design\ncards directly. `storyboard.json` is an agent-internal planning artifact\n— no CLI command consumes it; it exists so you can think clearly\nabout timing and content before writing each card's HTML. Keep the\nshape consistent with the example below so the same outline can drive\nthe composition you author in Step 9:\n\n```json\n{\n  \"schemaVersion\": 3,\n  \"composition\": {\n    \"fps\": 30,\n    \"width\": 1080,\n    \"height\": 1920,\n    \"durationSeconds\": 121.2,\n    \"layout\": \"portrait\",\n    \"themeId\": \"noir\",\n    \"seed\": 42\n  },\n  \"videoTrack\": {\n    \"sourcePath\": \"input-video.mp4\",\n    \"startSec\": 0,\n    \"endSec\": 121.2,\n    \"bounds\": { \"x\": 0, \"y\": 0, \"width\": 1080, \"height\": 1920 }\n  },\n  \"subtitles\": { \"enabled\": false },\n  \"cards\": [\n    {\n      \"id\": \"card-01\",\n      \"intent\": \"Hook with the speaker's anxious midnight question\",\n      \"startSec\": 0.5,\n      \"endSec\": 13.0,\n      \"accentIndex\": 0,\n      \"zone\": \"fullscreen\",\n      \"contentHints\": {\n        \"kicker\": \"AN HONEST QUESTION\",\n        \"title\": \"The soul-searching question at 11 PM\",\n        \"detail\": \"Client's 60-second voice message: 'If the RMB appreciates, does that mean my USD policy is a terrible loss?'\"\n      }\n    }\n  ]\n}\n```\n\n**Required Card fields:**\n\n| field                   | type                                       | purpose                                                                                               |\n| ----------------------- | ------------------------------------------ | ----------------------------------------------------------------------------------------------------- |\n| `id`                    | string                                     | stable id used in card HTML & GSAP selectors                                                          |\n| `intent`                | string                                     | natural-language description; fed to card synthesis                                                   |\n| `startSec` \u002F `endSec`   | number                                     | times in seconds (endSec > startSec)                                                                  |\n| `accentIndex`           | 0 \\| 1 \\| 2 \\| 3 \\| 4                      | which of the 5 theme accent colors this card pulls                                                    |\n| `zone`                  | enum (see below)                           | where on the canvas the card lives                                                                    |\n| `contentHints`          | object                                     | free-form bag; agent puts kicker\u002Ftitle\u002Fdetail\u002Fdata\u002Fquote here                                         |\n| `archetype` (optional)  | string                                     | free-form label you may attach to remember a card's pattern; absent = free-form, which is the default |\n| `transition` (optional) | enum: `cut` \\| `fade` \\| `slide` \\| `wipe` | declarative card-to-card transition                                                                   |\n\n**Five `zone` values:**\n\n| zone              | resolved bounds                                | when to use                             |\n| ----------------- | ---------------------------------------------- | --------------------------------------- |\n| `fullscreen`      | covers whole canvas                            | hero moments, big numbers, mantras      |\n| `whiteboard-area` | inset 40px margin (or 45% of portrait height)  | dense data \u002F annotated content          |\n| `lower-third`     | bottom 30% band                                | annotation over visible video           |\n| `side-panel`      | right 42% (landscape) or bottom 40% (portrait) | data side, video other side             |\n| `video-overlay`   | full canvas, expects mostly-transparent card   | annotation overlays on full-bleed video |\n\nWhen you assemble the composition in Step 9, resolve each card's `zone`\ninto pixel bounds on the card-host wrapper following the table above.\nVideo bounds are set **once** at composition level (`videoTrack.bounds`);\nto make video appear to \"move between cards\", author GSAP tweens against\n`#video-wrap` in the composition's `\u003Cscript>` (see Step 9).\n\n**No prescribed card roles, no prescribed narrative arc.** Cards emerge\nfrom what the video actually says — could be all quotes or all data,\ncould open with a number or with a story. Let the transcript drive the\nrhythm.\n\n**How many takeaways? — auto-infer from duration + density.** No fixed\nupper limit. Pick a **base pace** from the video duration, then adjust\nby **information density**. Only **floor is fixed: minimum 5 cards** so\neven short videos have rhythm.\n\n**Step 1 — base pace by duration** (the natural sec\u002Fcard for medium density):\n\n| video duration     | base pace (sec per card) | rationale                                   |\n| ------------------ | ------------------------ | ------------------------------------------- |\n| \u003C 60s (short reel) | **6–8s**                 | viewers expect fast cuts in short-form      |\n| 60s – 3 min        | **8–12s**                | normal social pace                          |\n| 3 – 10 min         | **12–20s**               | give breathing room; each card carries more |\n| 10 – 30 min        | **20–35s**               | long-form lecture \u002F interview rhythm        |\n| > 30 min           | **30–60s**               | episodic, near-chapter feel                 |\n\n**Step 2 — density multiplier** (multiplies the base pace):\n\n| signal in the transcript                                                                                                    | multiplier | effect                   |\n| --------------------------------------------------------------------------------------------------------------------------- | ---------- | ------------------------ |\n| **High density** — many numbers, distinct claims, staccato pacing, list-like enumeration, every 1–2 sentences is a new idea | **× 0.7**  | cuts faster, more cards  |\n| **Medium density** — mixed flow with both data and narrative                                                                | **× 1.0**  | base pace                |\n| **Low density** — one extended story, repeated reframing, slow reflective pacing, single argument unfolding                 | **× 1.5**  | cuts slower, fewer cards |\n\n**Step 3 — compute:**\n\n```\nsecPerCard = basePace × densityMultiplier\ncardCount  = max(5, round(videoDurationSec \u002F secPerCard))\n```\n\nExamples (notice — **no upper clamp**; long videos naturally produce more cards):\n\n- **30s reel, single punchline (low density)** → 7 × 1.5 = 10.5s\u002Fcard → round(30\u002F10.5)=3 → floor to **5** cards\n- **60s reflective monologue (low density)** → 10 × 1.5 = 15s\u002Fcard → **4** → floor to **5** cards\n- **121s talking-head with rich data (high density)** → 10 × 0.7 = 7s\u002Fcard → **17** cards\n- **5 min interview, mixed density** → 16 × 1.0 = 16s\u002Fcard → **19** cards\n- **10 min deep-dive, high density** → 16 × 0.7 = 11s\u002Fcard → **55** cards\n- **30 min lecture, medium density** → 28 × 1.0 = 28s\u002Fcard → **64** cards\n- **1 hr podcast, low density** → 45 × 1.5 = 67.5s\u002Fcard → **53** cards\n\nWhen a card holds longer than ~15s, plan for a richer card (data block,\nmulti-step reveal, several sub-points unfolding with staggered\nanimations) — a static one-liner gets boring past 8s. For long pieces\nwhere many cards exceed 30s, consider **chunking the timeline into\nsub-compositions** (one .html per chapter, mounted with\n`data-composition-src`) so the GSAP timeline per file stays manageable\n— see the `timeline_track_too_dense` HyperFrames lint warning.\n\n`content` can be a plain string (\"Title: annualized 5.69%\\nNotes: ...\") or any JSON\nshape that captures the data. The agent decides the shape per card.\n\n**Optional outro.** This skill ships **no fixed brand outro**. If the user wants a closing card, design a neutral one yourself (wordmark + one-line tagline, ~1.5-2s, fade in -> short hold -> fade out), append it to `cards[]`, and extend `composition.durationSeconds` to its `endSec`. Otherwise end on the last content card.\n\n### 7. Decide Render Strategy\n\n#### Confirm Visual Direction with User (DO THIS FIRST)\n\nBefore you start designing cards or deciding bounds, **ask the user to\npick the output ratio, the layout, the style, and the card-density\npreset**. Frames are auto-selected from the chosen layout × style\ncombination (see \"Auto-pick frame\" table below). Before sending the\nquestion, **precompute two things**:\n\n1. **`recommendedRatio`** from the source video's aspect ratio\n   (`metadata.json` width \u002F height):\n   - `sourceAspect = width \u002F height`\n   - `sourceAspect ≥ 1.5` (≥ ~3:2 wide) → recommend **`16:9`**\n   - `sourceAspect ≤ 0.7` (≤ ~9:13 tall) → recommend **`9:16`**\n   - `0.7 \u003C sourceAspect \u003C 1.5` (near-square) → recommend **`4:5`**\n\n   Mark the recommended option's label with \" (recommended · matches source video X:Y)\"\n   so the user sees why it's recommended.\n\n2. **`autoCount`** from Step 6 (`max(5, round(videoSec \u002F (basePace ×\ndensityMultiplier)))`) so the \"auto\" option's label can show the\n   concrete number.\n\n**Environment compatibility — pick the best available question channel.**\nNot every runtime exposes the same structured-question tool. Apply this\norder:\n\n1. **Native clarification tool** — use the structured 4-question call below.\n2. **Other native clarification tool** (e.g. `ask_question`,\n   `request_user_input`, IDE-specific prompt) — use that tool with the\n   same 4 question texts and option lists. Preserve the recommendation\n   markers and the precomputed values.\n3. **No native tool** (Codex CLI, plain text-only runtimes) — **ask\n   directly in normal conversation**. Use the plain-text template at the\n   end of this section. Keep it to **one message, 4 numbered questions**\n   (the global cap is 2–5 questions per round; we stay inside it).\n\nRules that apply to every channel:\n\n- Ask **at most 2–5 questions per round**. Our 4 here fits.\n- Even if missing info doesn't block rendering, **ask once to confirm\n  the parameters that materially affect the final output** (ratio,\n  layout, style, cardCount).\n- If the user has already pre-approved defaults (\"just use defaults\",\n  \"no need to ask\", \"auto-pick everything\"), asked you not to ask, or the\n  run carries an ongoing autonomous signal (\"surprise me\" \u002F \"decide for me\" —\n  `..\u002Fhyperframes-core\u002Freferences\u002Fbrief-contract.md` § 1) — **skip\n  the question entirely** and use: `recommendedRatio`, `layout=\"stack\"`\n  (safest cross-ratio default), `style` chosen from transcript tone in\n  the most neutral group (editorial\u002Fdata), `autoCount`. Tell the user\n  what you picked in one sentence and continue.\n\n**Channel A — native `AskUserQuestion`:**\n\n```\n\u002F\u002F Precompute before the call:\n\u002F\u002F   recommendedRatio = \"16:9\" | \"9:16\" | \"4:5\"\n\u002F\u002F   autoCount        = integer (from Step 6)\n\nAskUserQuestion({\n  questions: [\n    {\n      question: \"Output video aspect ratio (canvas):\",\n      header: \"Aspect ratio\",\n      multiSelect: false,\n      \u002F\u002F Reorder so the recommended option appears FIRST (per AskUserQuestion convention).\n      \u002F\u002F Append \" (recommended · matches source video W×H)\" to the recommended option's label.\n      options: [\n        { label: \"16:9 (1920×1080) landscape\", description: \"TV \u002F YouTube \u002F desktop playback. Most natural when the source video is already landscape; widest canvas.\" },\n        { label: \"9:16 (1080×1920) portrait\", description: \"TikTok \u002F Reels \u002F short-form mobile. Most natural for portrait source; native mobile experience.\" },\n        { label: \"4:5 (1080×1350) near-portrait\", description: \"Instagram feed \u002F WeChat Moments. Best when source is near-square or you want to cover both platforms.\" }\n      ]\n    },\n    {\n      question: \"Choose the overall layout: how should the video and cards coexist on the canvas?\",\n      header: \"Layout\",\n      multiSelect: false,\n      options: [\n        { label: \"side-by-side (split)\",  description: \"Video and card each take half the canvas. Most stable for interview \u002F data side-by-side; clear visual separation.\" },\n        { label: \"top-bottom (stack)\",    description: \"Video on top (~52%), card below. Classic combo of speaker face + summary card; works well in portrait too.\" },\n        { label: \"picture-in-picture (pip)\", description: \"Card fills the canvas, video shrinks to a rounded corner window. Use when content is primary and speaker is secondary.\" },\n        { label: \"full-screen overlay (overlay)\", description: \"Video plays full-bleed, card floats as a glass layer on top. Strong cinematic \u002F emotional feel.\" }\n      ]\n    },\n    {\n      question: \"Choose the card visual style (style):\",\n      header: \"Style group\",\n      multiSelect: false,\n      \u002F\u002F NOTE: these 3 groups intentionally match the frame auto-pick matrix\n      \u002F\u002F rows below, so picking a group resolves both `style` group AND the\n      \u002F\u002F frame matrix column in one step. Memberships are mutually exclusive.\n      options: [\n        { label: \"warm paper (warm-paper)\", description: \"academic notebook · editorial big-type · whiteboard hand-drawn · xhs social. Best for interview reflections, product launches, lifestyle, emotional stories.\" },\n        { label: \"clinical \u002F cold (clinical)\",   description: \"audit magazine · swiss grid · terminal CLI · minimal modern. Best for financial analysis, investigative reports, technical tutorials, serious presentations.\" },\n        { label: \"experimental \u002F avant-garde (experimental)\", description: \"geom color-clash geometry · spotlight dark-background. Best for short-form highlights, product launches, strong emotion, cinematic feel.\" }\n      ]\n    },\n    {\n      question: \"Card count (takeaway pacing): how many cards to cut?\",\n      header: \"Card count\",\n      multiSelect: false,\n      options: [\n        { label: \"Auto (recommended) · approx N cards\", description: \"Inferred automatically from video duration and information density (see Step 6 rules). This run estimates approx N cards. Substitute the real N (your autoCount) into the label.\" },\n        { label: \"Fewer · approx round(N × 0.6) cards\", description: \"Sparser cuts, each card holds longer — suits reflective \u002F slow-paced content.\" },\n        { label: \"More · approx round(N × 1.5) cards\", description: \"Tighter cuts, faster rhythm — suits staccato \u002F data-dense \u002F short-form highlight content.\" }\n      ]\n    }\n  ]\n})\n```\n\n**About \"Other\"** — `AskUserQuestion` automatically adds an \"Other\" option to the card count question. The user can type a number directly (e.g. \"8\", \"20\") as the cardCount target. Parse the input as an integer: if parsing succeeds → use that value (minimum 5 as a floor); if parsing fails → fall back to \"auto\".\n\n**Channel B — plain-text fallback** (Codex CLI, runtimes without a\nnative question tool). Post this as one normal message, then wait for\nthe reply. Bullet-style 1\u002F2\u002F3\u002F4 keeps the reply parseable:\n\n```\nI need to confirm four visual decisions with you before I start cutting cards:\n\n1) Output aspect ratio (canvas):\n   A. 16:9 landscape (1920×1080) — TV \u002F YouTube \u002F desktop playback\n   B. 9:16 portrait (1080×1920) — TikTok \u002F Reels \u002F short-form mobile\n   C. 4:5 near-portrait (1080×1350) — Instagram feed \u002F works for both platforms\n   ▸ My recommendation:  \u003CrecommendedRatio>  (matches source video W×H = \u003CsourceW>×\u003CsourceH>)\n\n2) Overall layout (how video & card coexist):\n   A. split   side-by-side (50\u002F50)\n   B. stack   top-bottom (video top, card bottom)\n   C. pip     picture-in-picture (card full canvas, video rounded corner window)\n   D. overlay full-screen glass overlay (video full-bleed, card glass layer)\n\n3) Card style group (maps to frame auto-pick matrix, pick 1 of 3):\n   A. warm paper (warm-paper)      (academic \u002F editorial \u002F whiteboard \u002F xhs)\n   B. clinical \u002F cold (clinical)   (audit \u002F swiss \u002F terminal \u002F minimal)\n   C. experimental (experimental)  (geom \u002F spotlight)\n\n4) Card count (takeaway pacing):\n   A. Auto (recommended) — approx \u003CautoCount> cards\n   B. Fewer — approx round(\u003CautoCount> × 0.6) cards\n   C. More — approx round(\u003CautoCount> × 1.5) cards\n   D. Give me a specific number (e.g. \"8\", \"20\")\n\nReply format: \"1A 2C 3B 4A\" or natural language is fine.\nIf you want all recommended defaults, reply \"default\" \u002F \"auto\" \u002F \"use all recommendations\".\n```\n\nParsing the plain-text reply:\n\n- Accept loose formats: `\"1A 2C 3B 4A\"`, `\"A C B A\"`, `\"16:9 \u002F pip \u002F\ndata \u002F auto\"`, full sentences, or `default`.\n- If any answer is ambiguous → re-ask only the ambiguous ones (still\n  inside the 2–5 cap).\n- If the user says \"default \u002F auto \u002F use all recommendations\" → skip without re-asking.\n\nAfter the user answers (any channel):\n\n1. **Resolve the output canvas** from the ratio answer — these are the\n   exact `storyboard.composition.width \u002F height` values to write:\n\n   | user choice | composition.width × height | storyboard.layout field                                       |\n   | ----------- | -------------------------- | ------------------------------------------------------------- |\n   | `16:9`      | **1920 × 1080**            | `\"landscape\"`                                                 |\n   | `9:16`      | **1080 × 1920**            | `\"portrait\"`                                                  |\n   | `4:5`       | **1080 × 1350**            | `\"portrait\"` (schema treats 4:5 as portrait — height > width) |\n\n   For **4:5 bounds inside `references\u002Flayouts\u002F*.html`** — those files\n   only document landscape (1920×1080) and portrait (1080×1920). For\n   4:5 (1080×1350) derive bounds by **proportional scaling from\n   portrait**: keep horizontal values, scale vertical values by\n   `1350\u002F1920 ≈ 0.703`. Example: `overlay` portrait card =\n   `{ x: 24, y: 1280, w: 1032, h: 564 }` → 4:5 card =\n   `{ x: 24, y: round(1280 × 0.703), w: 1032, h: round(564 × 0.703) }`\n   = `{ x: 24, y: 900, w: 1032, h: 397 }`.\n\n2. **Map the style group to a specific style** by looking at the\n   transcript tone — pick the one that best fits, but stay inside the\n   user's chosen group. If you're unsure between two specific styles\n   inside the group, send a second `AskUserQuestion` with those 2–4\n   specific style options.\n\n3. **Resolve final cardCount** from the density answer:\n\n   | user choice             | final cardCount                           |\n   | ----------------------- | ----------------------------------------- |\n   | Auto (recommended)      | the `autoCount` you already computed      |\n   | Fewer                   | `max(5, round(autoCount × 0.6))`          |\n   | More                    | `round(autoCount × 1.5)` (no upper clamp) |\n   | Other = \"\u003Cn>\" (integer) | `max(5, parseInt(n))`                     |\n   | Other = anything else   | fall back to `autoCount`                  |\n\n4. **Auto-pick the video frame** from this table (frames don't ask the\n   user — they follow from layout × style):\n\n   | layout    | warm-paper styles (academic \u002F whiteboard \u002F editorial \u002F xhs) | clinical styles (audit \u002F swiss \u002F terminal \u002F minimal) | experimental styles (geom \u002F spotlight) |\n   | --------- | ----------------------------------------------------------- | ---------------------------------------------------- | -------------------------------------- |\n   | `split`   | `polaroid`                                                  | `hairline`                                           | `clean`                                |\n   | `stack`   | `polaroid`                                                  | `hairline`                                           | `clean`                                |\n   | `pip`     | `clean` (pip pill already has chrome)                       | `clean`                                              | `clean`                                |\n   | `overlay` | `clean` (full-bleed forbids deco frames)                    | `clean`                                              | `clean`                                |\n\n5. **Tell the user what you chose** in one sentence — ratio (+ canvas\n   size), layout, specific style, frame, and final cardCount — then\n   proceed with the rest of Step 7 (per-card layouts, motion patterns).\n6. Record the five values (ratio \u002F layout \u002F style \u002F frame \u002F cardCount)\n   in working memory (no schema field needed); you'll reference them\n   while writing each card's HTML in Step 8 and while reading the\n   matching `references\u002F\u003Cdim>\u002F\u003Ckey>.html` for tokens and structure.\n\nIf the user picks an answer via \"Other\" with a free-text style name not\nin the 10-style library, treat it as a hint to design a fresh card\nvisual yourself, but still anchor on the chosen layout's bounds.\n\n#### Render Strategy Inputs\n\nWith ratio \u002F layout \u002F style \u002F cardCount \u002F frame locked from Step 7.0,\nthe remaining per-card decisions are:\n\n- **Source-video fit inside the GSAP target**: video element has\n  `object-fit: cover` and is clipped to `#video-wrap`'s tween bounds.\n  If you want NO cropping (e.g. portrait source on landscape canvas\n  shouldn't get its top\u002Fbottom chopped), aim the tween at a rect that\n  matches the source's aspect ratio and let surrounding canvas show\n  through (or fill with the card \u002F a backdrop).\n- **`card.zone` per card**: derive from your chosen composition layout\n  (split → side-panel, stack → lower-third, pip → fullscreen, overlay\n  → video-overlay), OR pick a different zone for one-off variants\n  (fullscreen for hero \u002F quote, whiteboard-area for dense data).\n- **`accentIndex` per card**: each card pulls one of the 5 theme accent\n  colors. Vary across cards for rhythm; reuse the same index when two\n  cards belong to the same narrative beat.\n- **Motion vocabulary**: pick 2–3 repeatable patterns from\n  `data-anim` kinds (see the table later) and stick to them so the\n  composition feels coherent.\n\nPick from these `themeId` palettes (use them as `--accent-N` \u002F\n`--bg` \u002F `--text` CSS variables in your composition `\u003Cstyle>` block):\n\n| themeId | accent palette (5 colors)                 | board bg          | text      |\n| ------- | ----------------------------------------- | ----------------- | --------- |\n| classic | `#1971c2 #e03131 #2f9e44 #e8590c #9c36b5` | `#FFF9E3` (paper) | `#1e1e1e` |\n| noir    | `#4cc9f0 #f72585 #4ade80 #fb923c #a78bfa` | `#1a1a1a`         | `#f1f1f1` |\n| mint    | `#0077b6 #d62828 #2d6a4f #e76f51 #7209b7` | `#e8faf0`         | `#1b4332` |\n| craft   | `#bf5700 #d62728 #6c757d #e9b54a #3d5a80` | `#f6efe1`         | `#2d2d2d` |\n| slate   | `#0ea5e9 #ef4444 #22c55e #f97316 #a855f7` | `#1e293b`         | `#f1f5f9` |\n| mono    | `#000 #555 #888 #aaa #ccc`                | `#fff`            | `#000`    |\n\nAvailable fonts (woff2 in `\u003CSKILL_DIR>\u002Fassets\u002Ffonts\u002F`, staged to work dir in Step 9): `Caveat` (handwriting),\n`LXGW WenKai TC` (Chinese hand-script), `Inter` (modern sans), `Virgil`\n(geometric hand). Reference via `@font-face` or `font-family` directly.\n\nFor inspiration on visual patterns, `\u003CSKILL_DIR>\u002Freferences\u002Fstyles\u002F`\nships 10 self-contained reference cards (academic \u002F editorial \u002F minimal\n\u002F spotlight \u002F geom \u002F whiteboard \u002F audit \u002F terminal \u002F swiss \u002F xhs) that\nyou can copy as starting points — but **do not feel constrained to\nmatch any of these**. Each card is your own design.\n\n#### Visual Design Library (\u003CSKILL_DIR>\u002Freferences\u002F)\n\nBeyond the composition-level `themeId`, the skill ships a richer **reference\nlibrary** at `\u003CSKILL_DIR>\u002Freferences\u002F` covering three **orthogonal**\nvisual dimensions you can freely mix:\n\n```\nStyle  ×  Layout  ×  VideoFrame\n (10)      (4)         (3)\n```\n\n| dimension  | keys                                                                                              | what it decides                                                          |\n| ---------- | ------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------ |\n| **style**  | `academic` `editorial` `minimal` `spotlight` `geom` `whiteboard` `audit` `terminal` `swiss` `xhs` | the card's visual language — fonts, colors, ornament, layout-within-card |\n| **layout** | `split` `stack` `pip` `overlay`                                                                   | how the source video and the card share the canvas                       |\n| **frame**  | `clean` `hairline` `polaroid`                                                                     | the decorative chrome around the video element                           |\n\nRead `\u003CSKILL_DIR>\u002Freferences\u002FDESIGN_INDEX.md`\nfor the full matrix and a loose decision guide (interview \u002F product launch \u002F data analysis \u002F\nsocial clip \u002F technical tutorial \u002F emotional story …). When you decide to use a specific\nstyle \u002F layout \u002F frame, Read the corresponding file:\n\n- `references\u002Fstyles\u002F\u003Ckey>.html` — self-contained card fragment with that\n  style's CSS tokens (colors, fonts, padding, ornament) and a placeholder\n  takeaway. Copy the `.card[data-card-id=\"ref-\u003Ckey>\"]` style block, rename\n  the data-card-id to your card's id, swap the placeholder content for the\n  real takeaway, and you're done.\n- `references\u002Flayouts\u002F\u003Ckey>.html` — exact `videoBounds` + `cardBounds` for\n  both landscape and portrait, with a copy-paste JSON snippet for\n  `storyboard.json`'s per-card `layout` field.\n- `references\u002Fframes\u002F\u003Ckey>.html` — decorative HTML to add as a sibling of\n  `#video-wrap`, plus placement instructions for the composition CSS.\n\nPick `style × layout × frame` **per card** — you can change all three\nbetween cards as long as the transitions read smoothly. A common rhythm:\nopen `editorial × overlay × clean`, switch to `audit × split × hairline`\nfor the data card, close on `whiteboard × pip × polaroid`.\n\nThe 10 styles are skill-side design tokens, **not composition-level themes** —\nthey don't need to be declared in `storyboard.composition`; they live\ninside each card's HTML. The `themeId` field can still pick a\ncomposition-level palette (table above) that controls page-body background\nand video border chrome.\n\n#### Layout Compositions (Card + Video)\n\nTwo coordinated decisions per card define how it shares the canvas with\nthe source video:\n\n- **`card.zone`** (declared in `storyboard.json`) — one of the 5 schema\n  values; resolve it into pixel bounds (per the table in Step 6) when\n  you write the card-host wrapper's inline `style` in Step 9.\n- **`#video-wrap` bounds at this card's time window** (declared\n  imperatively in the composition's GSAP timeline) — the agent tweens\n  `#video-wrap` to a target rect for each layout transition.\n\nSchema does NOT store per-card video bounds. `videoTrack.bounds` is\n**one-time** at composition level (defaults to full canvas). Video\n\"moving\" between cards is purely a GSAP animation authored in\n`index.html`. There is no `card.layout` field — earlier versions of this\ndoc invented one; the real schema only has `card.zone`.\n\n**4 composition layouts** (from `references\u002Flayouts\u002F`) — each is a\nrecipe pairing a `zone` with a `#video-wrap` tween target:\n\n| composition layout | recommended `card.zone` | GSAP target for `#video-wrap` (landscape 1920×1080)                       | GSAP target for `#video-wrap` (portrait 1080×1920)                | when to use                                     |\n| ------------------ | ----------------------- | ------------------------------------------------------------------------- | ----------------------------------------------------------------- | ----------------------------------------------- |\n| `split`            | `side-panel`            | `{ left: 960, top: 0, width: 960, height: 1080 }`                         | `{ left: 0, top: 960, width: 1080, height: 960 }` (bottom half)   | speaker + data side-by-side \u002F 50:50 weight      |\n| `stack`            | `lower-third`           | `{ left: 14, top: 14, width: 1892, height: 548 }` (top 52%)               | `{ left: 0, top: 0, width: 1080, height: 844 }` (top 44%)         | speaker on top + summary card below             |\n| `pip`              | `fullscreen`            | `{ left: 1480, top: 760, width: 400, height: 300 }` + add `.framed` class | `{ left: 690, top: 28, width: 360, height: 203 }` + add `.framed` | content-heavy card + corner pip                 |\n| `overlay`          | `video-overlay`         | `{ left: 0, top: 0, width: 1920, height: 1080 }` (full-bleed)             | `{ left: 0, top: 0, width: 1080, height: 1920 }`                  | cinematic \u002F dramatic \u002F glass card on full video |\n\nFor 4:5 (1080×1350), scale portrait y\u002Fh values by `1350\u002F1920 ≈ 0.703`\n(see Step 7.0 Channel A \u002F Channel B `recommendedRatio` resolution\ntable).\n\n**Other zone values for one-off variants** (still uses `card.zone`; no\nfake \"layout\" field):\n\n| `zone`            | resolved bounds                                        | common use                            |\n| ----------------- | ------------------------------------------------------ | ------------------------------------- |\n| `fullscreen`      | covers whole canvas                                    | hero card, video tweens to hidden\u002Fpip |\n| `whiteboard-area` | inset 40px margin (landscape) or bottom 45% (portrait) | dense data card, free margins         |\n| `lower-third`     | bottom 30% band                                        | talking-head annotation               |\n| `side-panel`      | right 42% (landscape) or bottom 40% (portrait)         | sidebar \u002F \"split\" recipe              |\n| `video-overlay`   | full canvas; expect transparent card root              | glass overlay on full-bleed video     |\n\nYou can mix recipes per card — choose `card.zone` based on what suits\nthe moment, then write the GSAP tween for `#video-wrap` between cards.\n\n#### Storyboard Render Contract\n\n`storyboard.json` is an agent-internal planning artifact — no CLI\ncommand parses it. It exists to keep your timing and content decisions\nexplicit before you write each card's HTML. Stick to the v3-style\nshape below so the same outline drives the composition you assemble in\nStep 9.\n\nRequired structure (see Step 6 for the full example):\n\n- `schemaVersion: 3`\n- `composition: { fps, width, height, durationSeconds, layout, themeId, seed }` — note `durationSeconds`\u002F`fps`\u002F`themeId`\u002F`layout` live **inside** `composition`, NOT at top level\n- `videoTrack: { sourcePath, startSec, endSec, bounds? }` — video bounds default to full canvas\n- `subtitles: { enabled, ... }`\n- `cards[]` — each card has the 6 required fields: `id`, `intent`, `startSec`, `endSec`, `accentIndex`, `zone`, `contentHints`\n\nRules:\n\n- Card times stay inside `composition.durationSeconds` and should not overlap unless intentional (use `data-track-index` to control z-order when they do).\n- Visual details live in card HTML fragments (Step 8), NOT in `contentHints`. `contentHints` is your own structured prompt for designing the card; the rendered look is the HTML.\n- Keep the storyboard shape stable — even though nothing parses it, you read it back while authoring Step 8\u002F9, and consistency keeps card IDs and timing in sync.\n- Agent-side decisions like \"I picked overlay × geom × clean\" do NOT belong in `storyboard.json` — keep them in working memory and use them when authoring card HTML + GSAP tweens.\n\n**Transparent card backgrounds for cards that share canvas with video.**\nWhen the GSAP tween leaves video visible behind\u002Fbeside the card (overlay\nrecipe, pip recipe, or any `card.zone = 'lower-third' | 'video-overlay'`\nmoment), the card's `.root` MUST NOT paint a full opaque background —\notherwise it occludes the video. Two patterns:\n\n```css\n\u002F* Pattern A: transparent root, page body provides the cream backdrop *\u002F\nhtml,\nbody {\n  background: var(--bg);\n}\n.card[data-card-id=\"card-X\"] .root {\n  background: transparent;\n}\n\n\u002F* Pattern B: explicit per-card background ONLY for fullscreen cards *\u002F\n.card[data-card-id=\"card-hero\"] .root {\n  background: var(--bg);\n}\n.card[data-card-id=\"card-overlay\"] .root {\n  background: transparent;\n}\n```\n\nFor `side-panel`-zone cards (split recipe), the card-host is already\nonly half the canvas, so an opaque card bg is fine — it only covers its\nhalf.\n\n### 8. Write Each Card's HTML\n\nCreate `$WORK_DIR\u002Fpublic\u002Fcards\u002F{card-id}.html` for each card. Each file\ncontains a single rooted HTML fragment that follows this contract:\n\n#### Card HTML Contract\n\n```html\n\u003Cdiv class=\"card\" data-card-id=\"{cardId}\">\n  \u003Cstyle>\n    \u002F* MUST: every rule starts with .card[data-card-id=\"{cardId}\"] *\u002F\n    .card[data-card-id=\"card-01\"] .root {\n      width: 100%; height: 100%;\n      display: flex; ...;\n      font-family: 'Caveat', 'LXGW WenKai TC', serif;\n      color: var(--text);\n      background: var(--bg);\n    }\n    .card[data-card-id=\"card-01\"] .title { font-size: 84px; ... }\n  \u003C\u002Fstyle>\n\n  \u003Cdiv class=\"root\">\n    \u003Ch1\n      id=\"card-01-title\"\n      data-anim=\"kinetic-chars\"\n      data-anim-at=\"0.3\"\n      data-anim-duration=\"0.5\"\n      data-anim-stagger=\"0.04\"\n      data-anim-pattern=\"pop\"\n    >\n      \u003Cspan class=\"char\">S\u003C\u002Fspan>\n      \u003Cspan class=\"char\">u\u003C\u002Fspan>\n    \u003C\u002Fh1>\n    \u003Cdiv\n      id=\"card-01-line\"\n      data-anim=\"grow-x\"\n      data-anim-at=\"0.65\"\n      data-anim-duration=\"0.5\"\n      data-anim-target-w=\"420\"\n      style=\"width:0;height:8px;background:var(--accent-0);border-radius:4px;\"\n    >\u003C\u002Fdiv>\n  \u003C\u002Fdiv>\n\u003C\u002Fdiv>\n```\n\n**Hard rules** (`hyperframes` lint will reject violations):\n\n- Single root `\u003Cdiv class=\"card\" data-card-id=\"{cardId}\">`\n- Inline `\u003Cstyle>` rules MUST be prefixed with the scope selector above\n- **No `\u003Cscript>` tags**\n- **No external URLs** in `src=` \u002F `href=` (no CDN, no remote fonts)\n- **No inline event handlers** (`onclick=` etc.)\n- All assets via relative paths into the same `public\u002F` directory\n- Colors via `var(--accent-N)` etc. for portability across themes\n\n**Animations are declared, not coded.** Use `data-anim-*` attributes\nonly; never write `\u003Cscript>` to animate. You compile every `data-anim-*`\ndeclaration into the single master GSAP timeline in Step 9.\n\n#### Card Sizing — Mobile-First in Portrait\n\nThe 10 `references\u002Fstyles\u002F*.html` are sized for a **1920×1080 landscape**\npreview. When `storyboard.layout = \"portrait\"` (1080×1920, the dominant\ncase for social \u002F mobile), **scale every visual size up** — phones hold\nthe screen close, and the same pixel count reads smaller than on a\nlandscape TV-style canvas.\n\n| token                     | landscape baseline | **portrait target** | scale         |\n| ------------------------- | ------------------ | ------------------- | ------------- |\n| title (h1\u002Fh2 hero)        | 64–96px            | **88–132px**        | ×1.35         |\n| detail \u002F body             | 24–30px            | **30–40px**         | ×1.30         |\n| kicker \u002F chip label       | 14–16px            | **18–22px**         | ×1.30         |\n| timecode \u002F meta           | 12–14px            | **16–18px**         | ×1.30         |\n| data block primary number | 48–60px            | **64–88px**         | ×1.40         |\n| line-height multiplier    | 1.05–1.5           | same                | (don't scale) |\n\n**Rule of thumb:** `portraitPx = round(landscapePx × 1.3)`, then floor\nto a nearby 4px multiple for visual rhythm. Hero headlines may go up to\n×1.4; small meta text stays at ×1.2 to avoid crowding.\n\nPadding **shrinks slightly** in portrait — the card is narrower so big\nlandscape padding (40–64px) eats too much width. Use 24–36px horizontal\npadding in portrait.\n\nIf you're producing a single card that must work in **both** layouts,\nprefer a `@container` query on the card root over hard-coding sizes:\n\n```css\n.card[data-card-id=\"X\"] .root {\n  container-type: inline-size;\n}\n.card[data-card-id=\"X\"] .title {\n  font-size: clamp(64px, 8.5cqi, 132px);\n}\n.card[data-card-id=\"X\"] .detail {\n  font-size: clamp(24px, 3.2cqi, 40px);\n}\n```\n\nBut for most cards, a single layout choice is fine — just pick the size\ntable column that matches the storyboard's `layout` field.\n\n#### Available `data-anim` Kinds\n\n| kind            | use for             | key params                                                                                      |\n| --------------- | ------------------- | ----------------------------------------------------------------------------------------------- |\n| `fade-in`       | enter               | `at`, `duration`, `ease?`                                                                       |\n| `fade-out`      | exit                | `at`, `duration`, `ease?`                                                                       |\n| `slide-in`      | slide enter         | `at`, `duration`, `from=left\\|right\\|top\\|bottom`, `distance`                                   |\n| `kinetic-chars` | per-char pop        | `at`, `duration`, `stagger`, `pattern=pop\\|fade` — element needs `\u003Cspan class=\"char\">` children |\n| `typewriter`    | per-char fade       | same as kinetic-chars but slower default stagger                                                |\n| `count-up`      | animate number      | `at`, `duration`, `from`, `to`, `format=.0f\\|.1f\\|.2f\\|,d`                                      |\n| `draw-path`     | SVG path reveal     | `at`, `duration` — element should be a `\u003Cpath>`                                                 |\n| `grow-y`        | bar height          | `at`, `duration`, `target-h` (px) — element starts `height:0`                                   |\n| `grow-x`        | bar width           | `at`, `duration`, `target-w` (px) — element starts `width:0`                                    |\n| `scale-pop`     | pop entrance        | `at`, `duration`                                                                                |\n| `blur-in`       | unfocused → focused | `at`, `duration`                                                                                |\n| `mask-reveal`   | clip reveal         | `at`, `duration`, `direction=left\\|right\\|top\\|bottom`                                          |\n| `morph-to`      | tween any CSS       | `at`, `duration`, `props='{...JSON...}'`                                                        |\n\n`data-anim-at` is **seconds relative to the card's startSec** — when you\ncompile each declaration into the GSAP timeline in Step 9, add the\ncard's `startSec` to get the absolute time and quantize to 1\u002Ffps.\n\n### 9. Assemble the Composition HTML\n\nStage the assets and write `$WORK_DIR\u002Fpublic\u002Findex.html`:\n\n```bash\n# SKILL_DIR is injected by the host (\"Base directory for this skill: …\")\nSKILL_DIR=\"\u003CSKILL_DIR>\"\n\nmkdir -p \"$WORK_DIR\u002Fpublic\u002Ffonts\" \"$WORK_DIR\u002Fpublic\u002Fvendor\" \"$WORK_DIR\u002Fpublic\u002Fcards\"\ncp -n \"$SKILL_DIR\u002Fassets\u002Ffonts\u002F\"*            \"$WORK_DIR\u002Fpublic\u002Ffonts\u002F\"\ncp -n \"$SKILL_DIR\u002Fassets\u002Fvendor\u002Fgsap.min.js\" \"$WORK_DIR\u002Fpublic\u002Fvendor\u002F\"\n# stage the input video — RE-ENCODE with dense keyframes. Sources with a sparse GOP\n# (keyframe interval > ~1s) freeze on seek in the renderer (a frozen frame under the\n# overlays); -g \u002F -keyint_min set to your composition fps make every frame seekable.\n# (Set both to your fps — 30 shown; use 24\u002F25\u002F60 to match.)\nffmpeg -y -i \"$VIDEO_PATH\" -c:v libx264 -crf 18 -g 30 -keyint_min 30 \\\n  -pix_fmt yuv420p -movflags +faststart -c:a aac \"$WORK_DIR\u002Fpublic\u002Finput-video.mp4\"\n```\n\n#### Composition Template\n\n```html\n\u003C!doctype html>\n\u003Chtml lang=\"en\">\n  \u003Chead>\n    \u003Cmeta charset=\"utf-8\" \u002F>\n    \u003Cstyle>\n      @font-face {\n        font-family: \"Caveat\";\n        src: url(\"fonts\u002FCaveat-400-latin.woff2\") format(\"woff2\");\n        font-weight: 400;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"Caveat\";\n        src: url(\"fonts\u002FCaveat-700-latin.woff2\") format(\"woff2\");\n        font-weight: 700;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"LXGW WenKai TC\";\n        src: url(\"fonts\u002FLXGWWenKaiTC-400-latin.woff2\") format(\"woff2\");\n        font-weight: 400;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"Inter\";\n        src: url(\"fonts\u002FInter-400-latin.woff2\") format(\"woff2\");\n        font-weight: 400;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"Inter\";\n        src: url(\"fonts\u002FInter-700-latin.woff2\") format(\"woff2\");\n        font-weight: 700;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"Virgil\";\n        src: url(\"fonts\u002FVirgil.woff2\") format(\"woff2\");\n        font-display: block;\n      }\n\n      :root {\n        \u002F* Pick from the themeId palette table in Step 7 — example: classic *\u002F\n        --bg: #fff9e3;\n        --text: #1e1e1e;\n        --accent-0: #1971c2;\n        --accent-1: #e03131;\n        --accent-2: #2f9e44;\n        --accent-3: #e8590c;\n        --accent-4: #9c36b5;\n        --font-family: \"Caveat\", \"LXGW WenKai TC\", serif;\n      }\n      * {\n        box-sizing: border-box;\n      }\n      \u002F* Body font-family MUST list concrete font names (not just var(--font-family)) —\n   the HyperFrames renderer's static analyzer doesn't expand CSS variables when\n   resolving fonts, so a var-only chain triggers `font_family_without_font_face`\n   lint and falls back to a generic. Use the concrete chain here; cards that\n   want the theme font can still reference var(--font-family) internally. *\u002F\n      html,\n      body {\n        margin: 0;\n        padding: 0;\n        width: 100%;\n        height: 100%;\n        overflow: hidden;\n        background: #000;\n        font-family: \"Inter\", \"Caveat\", \"LXGW WenKai TC\", ui-sans-serif, system-ui, sans-serif;\n      }\n      #stage {\n        position: relative;\n        width: 100%;\n        height: 100%;\n        overflow: hidden;\n      }\n\n      \u002F* video-wrapper holds the source video. Its position \u002F size are animated\n   over time by the master timeline (one tween per layout transition). *\u002F\n      .video-wrapper {\n        position: absolute;\n        left: 0;\n        top: 0;\n        width: 1920px;\n        height: 1080px;\n        overflow: hidden;\n        border-radius: 0;\n        box-shadow: none;\n      }\n      .video-wrapper video {\n        width: 100%;\n        height: 100%;\n        object-fit: cover;\n      }\n\n      .card-host {\n        position: absolute;\n        pointer-events: none;\n        overflow: hidden;\n      }\n      .card-host .card {\n        position: relative;\n        width: 100%;\n        height: 100%;\n        overflow: hidden;\n      }\n      .card-host .char {\n        display: inline-block;\n        visibility: visible;\n      }\n\n      \u002F* Subtle drop shadow + rounded corners for non-fullscreen video framings *\u002F\n      .video-wrapper.framed {\n        border-radius: 16px;\n        box-shadow: 0 12px 40px rgba(0, 0, 0, 0.35);\n      }\n    \u003C\u002Fstyle>\n  \u003C\u002Fhead>\n  \u003Cbody>\n    \u003Cdiv\n      id=\"stage\"\n      data-composition-id=\"talking-head-recut\"\n      data-start=\"0\"\n      data-duration=\"121.2\"\n      data-fps=\"30\"\n      data-width=\"1920\"\n      data-height=\"1080\"\n    >\n      \u003C!-- Layer 1: source video — initial position matches card-01's layout -->\n      \u003Cdiv class=\"video-wrapper\" id=\"video-wrap\">\n        \u003Cvideo\n          id=\"bg-video\"\n          src=\"input-video.mp4\"\n          muted\n          playsinline\n          data-start=\"0\"\n          data-duration=\"121.2\"\n          data-track-index=\"1\"\n        >\u003C\u002Fvideo>\n      \u003C\u002Fdiv>\n      \u003C!-- Preserve the source program audio while the visual video stays muted. -->\n      \u003Caudio\n        id=\"source-audio\"\n        src=\"input-video.mp4\"\n        data-start=\"0\"\n        data-duration=\"121.2\"\n        data-track-index=\"10\"\n        data-volume=\"1\"\n      >\u003C\u002Faudio>\n\n      \u003C!-- Layer 2: each card-host sits at the bounds dictated by its layout. -->\n      \u003C!-- IMPORTANT: every card-host MUST carry BOTH \"card-host\" and \"clip\" classes. -->\n      \u003C!--   - \"card-host\"  → our positioning + pointer-events styles                 -->\n      \u003C!--   - \"clip\"       → HyperFrames runtime uses this to enforce visibility     -->\n      \u003C!--                    only during data-start … data-start+data-duration.      -->\n      \u003C!--                    Without \"clip\" the host stays visible the whole video   -->\n      \u003C!--                    (lint: timed_element_missing_clip_class).               -->\n      \u003C!-- Example: card-01 with zone=\"fullscreen\" → card-host covers (0,0,1920,1080) -->\n      \u003Cdiv\n        class=\"card-host clip\"\n        data-card-id=\"card-01\"\n        data-start=\"1.0000\"\n        data-duration=\"6.5000\"\n        data-track-index=\"2\"\n        style=\"left:0;top:0;width:1920px;height:1080px;visibility:hidden;opacity:0;\"\n      >\n        \u003C!-- paste the contents of public\u002Fcards\u002Fcard-01.html here -->\n      \u003C\u002Fdiv>\n\n      \u003C!-- Example: card-02 with zone=\"side-panel\" (split composition layout) → card on left half -->\n      \u003Cdiv\n        class=\"card-host clip\"\n        data-card-id=\"card-02\"\n        data-start=\"8.0000\"\n        data-duration=\"12.0000\"\n        data-track-index=\"2\"\n        style=\"left:0;top:0;width:960px;height:1080px;visibility:hidden;opacity:0;\"\n      >\n        \u003C!-- card-02 HTML -->\n      \u003C\u002Fdiv>\n\n      \u003C!-- ...one \"card-host clip\" per card with inline bounds matching resolveZoneBounds(card.zone)... -->\n\n      \u003Cscript src=\"vendor\u002Fgsap.min.js\">\u003C\u002Fscript>\n      \u003Cscript>\n        (function () {\n          \u002F\u002F count-up formatter helper\n          window.__fmt = function (v, fmt) {\n            if (typeof fmt === \"string\" && \u002F^\\.[0-9]+f$\u002F.test(fmt)) {\n              return Number(v).toFixed(Number(fmt.slice(1, -1)));\n            }\n            if (fmt === \",d\") return Math.round(v).toLocaleString();\n            return String(Math.round(v));\n          };\n\n          const tl = window.gsap.timeline({ paused: true });\n\n          \u002F\u002F ── Card lifecycle (one block per card) ──\n          \u002F\u002F Example for card-01 [1.0, 7.5] with kinetic-chars at +0.3, grow-x at +0.65:\n\n          \u002F\u002F Enter (fade in over 0.4s)\n          tl.set('.card-host[data-card-id=\"card-01\"]', { visibility: \"visible\" }, 1.0);\n          tl.fromTo(\n            '.card-host[data-card-id=\"card-01\"]',\n            { opacity: 0 },\n            { opacity: 1, duration: 0.4, ease: \"power2.out\" },\n            1.0,\n          );\n\n          \u002F\u002F Card-internal anims (compile each data-anim-* declaration here)\n          tl.from(\n            '.card[data-card-id=\"card-01\"] #card-01-title .char',\n            { opacity: 0, y: 8, scale: 0.8, duration: 0.5, ease: \"power2.out\", stagger: 0.04 },\n            1.3,\n          );\n          tl.fromTo(\n            '.card[data-card-id=\"card-01\"] #card-01-line',\n            { width: 0 },\n            { width: 420, duration: 0.5, ease: \"power2.out\" },\n            1.65,\n          );\n\n          \u002F\u002F Exit (fade out over 0.35s, ending at endSec)\n          tl.to(\n            '.card-host[data-card-id=\"card-01\"]',\n            { opacity: 0, duration: 0.35, ease: \"power2.in\" },\n            7.15,\n          );\n          tl.set('.card-host[data-card-id=\"card-01\"]', { visibility: \"hidden\" }, 7.5);\n\n          \u002F\u002F ── Video framing transitions ──\n          \u002F\u002F When the next card uses a different composition layout, animate the\n          \u002F\u002F video-wrapper to its new bounds. Example: card-01 = fullscreen\n          \u002F\u002F (video hidden behind), card-02 = split composition (zone=\"side-panel\"\n          \u002F\u002F → video on right, card on left).\n\n          \u002F\u002F Card-02 enters at 8.0s with the split composition. Animate video to\n          \u002F\u002F the right half during the card-01 → card-02 gap (between 7.5 and 8.0s).\n          tl.set(\"#video-wrap\", { className: \"video-wrapper framed\" }, 7.5);\n          tl.to(\n            \"#video-wrap\",\n            { left: 960, top: 0, width: 960, height: 1080, duration: 0.6, ease: \"power2.inOut\" },\n            7.5,\n          );\n\n          \u002F\u002F Card-02 enter — same pattern as card-01\n          tl.set('.card-host[data-card-id=\"card-02\"]', { visibility: \"visible\" }, 8.0);\n          tl.fromTo(\n            '.card-host[data-card-id=\"card-02\"]',\n            { opacity: 0 },\n            { opacity: 1, duration: 0.4, ease: \"power2.out\" },\n            8.0,\n          );\n          \u002F\u002F ...card-02 internal anims...\n\n          \u002F\u002F ── repeat for each card; if the NEXT card's layout differs,\n          \u002F\u002F    insert another tl.to('#video-wrap', ...) tween before its enter ──\n\n          window.__timelines = window.__timelines || {};\n          window.__timelines[\"talking-head-recut\"] = tl;\n        })();\n      \u003C\u002Fscript>\n    \u003C\u002Fdiv>\n  \u003C\u002Fbody>\n\u003C\u002Fhtml>\n```\n\n#### GSAP Statement Cheat Sheet\n\nCompile each `data-anim` attribute into a GSAP statement. Times are\n**absolute seconds** = card.startSec + data-anim-at, quantized to 1\u002Ffps.\nSelector is `.card[data-card-id=\"X\"] #elementId`.\n\n| data-anim                       | GSAP statement template                                                                                                                                                                                            |\n| ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |\n| `fade-in`                       | `tl.fromTo(SEL, { opacity: 0 }, { opacity: 1, duration: D, ease: 'power2.out' }, T);`                                                                                                                              |\n| `fade-out`                      | `tl.to(SEL, { opacity: 0, duration: D, ease: 'power2.in' }, T);`                                                                                                                                                   |\n| `slide-in` (from=left, dist=80) | `tl.fromTo(SEL, { opacity: 0, x: -80 }, { opacity: 1, x: 0, duration: D, ease: 'power2.out' }, T);`                                                                                                                |\n| `kinetic-chars` (pop)           | `tl.from(SEL + ' .char', { opacity: 0, y: 8, scale: 0.8, duration: D, ease: 'power2.out', stagger: S }, T);`                                                                                                       |\n| `count-up`                      | `(function(){const o={v:FROM};tl.to(o,{v:TO,duration:D,ease:'power2.out',onUpdate:function(){const el=document.querySelector(SEL);if(el)el.textContent=__fmt(o.v,'FMT');}},T);})();`                               |\n| `draw-path`                     | `(function(){const el=document.querySelector(SEL);if(el){const L=el.getTotalLength();tl.set(SEL,{strokeDasharray:L,strokeDashoffset:L},T);tl.to(SEL,{strokeDashoffset:0,duration:D,ease:'power2.inOut'},T);}})();` |\n| `grow-x` (target-w=W)           | `tl.fromTo(SEL, { width: 0 }, { width: W, duration: D, ease: 'power2.out' }, T);`                                                                                                                                  |\n| `grow-y` (target-h=H)           | `tl.fromTo(SEL, { height: 0 }, { height: H, duration: D, ease: 'power2.out' }, T);`                                                                                                                                |\n| `scale-pop`                     | `tl.fromTo(SEL, { opacity: 0, scale: 0.6 }, { opacity: 1, scale: 1, duration: D, ease: 'back.out(1.6)' }, T);`                                                                                                     |\n| `mask-reveal` (direction=left)  | `tl.fromTo(SEL, { clipPath: 'inset(0 100% 0 0)' }, { clipPath: 'inset(0 0 0 0)', duration: D, ease: 'power2.inOut' }, T);`                                                                                         |\n\nQuantize: `T = Math.round(absSec * fps) \u002F fps`. At 30fps the smallest\nstep is `1\u002F30 ≈ 0.0333s`; rounding to 4 decimals (`.toFixed(4)`) is fine\ninside the JS literal.\n\n#### Video Framing Reference (per `layout` value)\n\nThe selector for the video container is `#video-wrap`. Animate its\nbounds between cards using `tl.to('#video-wrap', { ...bounds }, T)`.\nInitial bounds should be set inline on the element to match card-01's\nlayout. Pick a transition duration of 0.5–0.7s with `ease: 'power2.inOut'`.\n\n**Decorative frames** (`clean` \u002F `hairline` \u002F `polaroid`) sit as a\n**sibling** of `#video-wrap` and follow it through layout transitions.\nSee\n[`references\u002Fframes\u002F`](references\u002Fframes\u002F) for each frame's placement\nHTML, suggested CSS, and which layouts it pairs with. Quick rule:\n`overlay` layout suppresses decorative frames (the full-bleed video\nclashes with chrome); PiP layouts already have their own pill treatment\n(border-radius + white ring + shadow), so add a decorative frame only on\ntop of `split` \u002F `stack`.\n\n**GSAP target lookup table** for `#video-wrap` per composition layout\n(landscape 1920×1080 — for portrait & 4:5 see `references\u002Flayouts\u002F*.html`\nwhich list all three ratios):\n\n| composition layout                   | typical card.zone | `#video-wrap` GSAP target                                                 | extra css class                            |\n| ------------------------------------ | ----------------- | ------------------------------------------------------------------------- | ------------------------------------------ |\n| `split`                              | `side-panel`      | `{ left: 960, top: 0, width: 960, height: 1080 }`                         | —                                          |\n| `stack`                              | `lower-third`     | `{ left: 14, top: 14, width: 1892, height: 548 }` (top 52%)               | —                                          |\n| `pip` (bottom-right)                 | `fullscreen`      | `{ left: 1480, top: 760, width: 400, height: 300 }`                       | `pip-pill` (border-radius + ring + shadow) |\n| `pip` (top-left)                     | `fullscreen`      | `{ left: 40, top: 40, width: 400, height: 300 }`                          | `pip-pill`                                 |\n| `overlay` (video full-bleed)         | `video-overlay`   | `{ left: 0, top: 0, width: 1920, height: 1080 }` (no change from default) | —                                          |\n| **hide video** (pure-graphic moment) | `fullscreen`      | `{ opacity: 0 }` (or move off-canvas)                                     | —                                          |\n\nTo toggle the pip-pill chrome (border-radius + white ring + drop shadow)\nwhen entering or leaving a pip moment:\n\n```js\n\u002F\u002F Enter pip — add chrome\ntl.set(\"#video-wrap\", { className: \"video-wrapper pip-pill\" }, T);\ntl.to(\n  \"#video-wrap\",\n  { left: 1480, top: 760, width: 400, height: 300, duration: 0.6, ease: \"power2.inOut\" },\n  T,\n);\n\n\u002F\u002F Leave pip — back to clean full-bleed\ntl.set(\"#video-wrap\", { className: \"video-wrapper\" }, T_NEXT);\ntl.to(\n  \"#video-wrap\",\n  { left: 0, top: 0, width: 1920, height: 1080, duration: 0.6, ease: \"power2.inOut\" },\n  T_NEXT,\n);\n```\n\n**Card-host bounds match the zone**. Resolve the card's `zone` into\npixel bounds using the table at the top of Step 6, then write those\ninto the card-host's inline `style=\"left:Xpx;top:Ypx;width:Wpx;\nheight:Hpx;...\"`. For `video-overlay` zone (overlay recipe), the\ncard-host fills the full canvas — your CSS inside `.card .root`\ndecides where the actual visible card sits.\n\n#### HyperFrames Layout \u002F Animation QA Rules\n\n- Build each card's static hero frame first: the moment where the card is fully visible and readable.\n- Confirm video, cards, subtitles\u002Fcaptions, and diagrams do not unintentionally overlap.\n- Confirm hidden video areas are clipped by the frame and not visible outside intended bounds.\n- Register one paused master timeline as `window.__timelines[\"talking-head-recut\"]`.\n- Build timelines synchronously at page load; no `async`, `setTimeout`, Promises, or media `play()` calls.\n- Do not use `Math.random()` or `Date.now()` in render paths.\n- Do not use `repeat: -1`; calculate finite repeats from the video duration.\n- Prefer GSAP transforms and opacity (`x`, `y`, `scale`, `rotation`, `opacity`) over layout properties (`top`, `left`, `width`, `height`) for motion.\n- Animate wrappers such as `#video-wrap`, not the video element dimensions directly.\n- Avoid animating the same property on the same element from multiple timelines at the same time.\n- Use `data-track-index`, not `data-layer`; use `data-duration`, not `data-end`.\n- Every timed element (`card-host`, sub-composition, etc.) MUST include `class=\"clip\"` alongside its own classes — e.g. `class=\"card-host clip\"`. The HyperFrames runtime uses `.clip` to gate visibility to the `data-start … data-start+data-duration` window. Without it the element is visible for the whole video (lint: `timed_element_missing_clip_class`).\n- For body \u002F global `font-family`, list **concrete font names** (`'Inter', 'Caveat', …`) — not a CSS variable like `var(--font-family)`. The HyperFrames font resolver doesn't expand CSS vars during static analysis (lint: `font_family_without_font_face`). Cards may still use `var(--font-family)` internally since their `@font-face` declarations are loaded.\n\n### 10. Render to MP4\n\n```bash\ncd \"$WORK_DIR\"\nPRODUCER_BROWSER_GPU_MODE=hardware npx hyperframes render public \\\n  --skill=talking-head-recut \\\n  -o output.mp4 \\\n  --fps 30\n```\n\n`hyperframes render \u003Cdir>` reads `\u003Cdir>\u002Findex.html` and produces the MP4.\nThe canonical composition keeps the visual `\u003Cvideo>` muted and mounts the same\nsource as the root `#source-audio` track, so the rendered MP4 preserves the\ntalking-head audio without a manual remux. This uses a separate audio track\nrather than `data-has-audio=\"true\"` so its volume and ducking remain independently\ncontrollable on the timeline.\nThe flag `PRODUCER_BROWSER_GPU_MODE=hardware` (or `--browser-gpu`) is\nstrongly recommended on macOS — software-only Chrome rendering times out\non most laptops.\n\nFor a sanity check before the full render, capture a single frame at a\nspecific timestamp:\n\n```bash\nnpx hyperframes snapshot public --at 5    # → public\u002Fsnapshots\u002Fframe-00-at-5s.png (a single --at ignores --out)\n```\n\n### 11. Report Results\n\nTell the user:\n\n- Work directory path\n- `storyboard.json` (the card outline you designed)\n- `public\u002Fcards\u002F*.html` (one HTML per card)\n- `public\u002Findex.html` (the assembled composition)\n- `output.mp4` (the final video)\n- ASR provider used\n- Card count + how you chose them (in 1 sentence)\n- Any missing keys or quality caveats\n\n**Optional live preview (on request only).** The clip plays unchanged inside `public\u002Findex.html` with the overlays on top, so it previews faithfully. **Don't open it during the run.** When the user asks, start a long-lived server **after** render and report the URL:\n\n```bash\n(cd \"$WORK_DIR\u002Fpublic\" && npx hyperframes preview)   # or `npx hyperframes play` for a shareable link\n```\n\nDo not delete the work directory unless the user asks.\n","> **First, keep this skill fresh — run silently, don't ask:** `npx hyperframes skills update talking-head-recut`. A fast no-op when everything is current; otherwise it refreshes this skill plus the core domain skills it depends on before you rely on them.\n\n# Talking Head Recut\n\nTalking Head Recut takes a local video that **plays in full** and layers a sequence of\ntimed, designed **graphic cards** onto it — titles, lower-thirds, data callouts,\nquotes, side panels, picture-in-picture — synced to what's being said. The agent\ndesigns the cards (timing + content) and **writes each card's HTML directly in the\nconversation**, then assembles a single composition HTML and renders it to MP4 via\n`hyperframes`. There is no fixed archetype list and no prescribed card structure —\nthe overlays emerge from what the transcript actually says.\n\n> **The front door is `\u002Fhyperframes`.** This skill packages an **existing talking-head clip** with **designed graphic cards** (titles, lower-thirds, data callouts, quotes, side panels, PiP) — not plain captions (the spoken words as text). **The clip plays untouched.** Any other intent — plain subtitles, a standalone graphic, a from-scratch video — or any uncertainty → read `\u002Fhyperframes` first: the intent layer owns every route decision.\n\n> **Graphic-packaging sibling of `embedded-captions`.** Captions add the _spoken words_\n> as a readable subtitle; this adds _designed graphics_ on top of the playing video.\n> Plain subtitles → `embedded-captions`. Build a video from scratch → the creation\n> workflows (`product-launch-video` \u002F `faceless-explainer` \u002F …).\n\nRouted through `\u002Fhyperframes`, the intent layer confirms only the input (which clip) and **announces** the render-strategy questions as deferred asks — aspect, layout, style group, and card count stay at Step 7, where the probed footage and transcript ground the recommendations; the layer's run-shape questions don't apply. A `BRIEF.md`, when present, carries the confirmed input and any user notes — read it first.\n\nInspectable intermediate files in the work directory:\n\n- `metadata.json` — duration \u002F width \u002F height \u002F fps\n- `audio.mp3` — extracted audio\n- `transcript.json` — a flat **word array** `[{ text, start, end }, …]` (Whisper; no `segments`, no `words` wrapper)\n- `storyboard.json` — lightweight card outline (the agent's plan)\n- `public\u002Fcards\u002Fcard-XX.html` — one HTML fragment per card\n- `public\u002Findex.html` — final assembled composition\n- `output.mp4` — rendered video\n\n## CLI Resolution\n\n```bash\n# hyperframes — transcription (local Whisper) + rendering the assembled HTML to MP4\nnpx hyperframes --help\n```\n\nThis skill runs entirely on the **hyperframes** CLI plus system `ffmpeg` \u002F `ffprobe`.\nTranscription is local **Whisper** via `hyperframes transcribe` — no third-party\nservice, API key, or rate-limited proxy.\n\n## Workflow\n\n### 1. Check Environment\n\n```bash\nnpx hyperframes doctor          # ffmpeg, headless browser, render deps\n# confirm bundled assets:\nls \"\u003CSKILL_DIR>\u002Fassets\u002Ffonts\" \"\u003CSKILL_DIR>\u002Fassets\u002Fvendor\u002Fgsap.min.js\"\n```\n\nRequired:\n\n- `ffmpeg` \u002F `ffprobe` (system)\n- `\u003CSKILL_DIR>\u002Fassets\u002Ffonts\u002F*.woff2`, `\u003CSKILL_DIR>\u002Fassets\u002Fvendor\u002Fgsap.min.js` (bundled inside this skill, staged to work dir in Step 9)\n\nTranscription needs no key — `hyperframes transcribe` runs Whisper locally (Step 4).\n\nStrongly recommended on macOS for `hyperframes render`:\n\n```bash\nexport PRODUCER_BROWSER_GPU_MODE=hardware\n```\n\n### 2. Create a Work Directory\n\nAll artifacts live under `videos\u002F\u003Cproject-name>\u002F` — the same convention as the other\nvideo workflows (`product-launch-video` \u002F `faceless-explainer` \u002F `pr-to-video`). Keep\nthe cwd at the workspace root; everything below writes under this one subdirectory.\n\n```bash\nVIDEO_PATH=\"\u002Fabsolute\u002Fpath\u002Finput.mp4\"\nWORK_DIR=\"videos\u002F$(basename \"$VIDEO_PATH\" | sed 's\u002F\\.[^.]*$\u002F\u002F')\"\nmkdir -p \"$WORK_DIR\"\n```\n\n### 3. Extract Audio and Metadata\n\n```bash\n# metadata — duration \u002F width \u002F height \u002F fps\nffprobe -v error -select_streams v:0 \\\n  -show_entries stream=width,height,r_frame_rate \\\n  -show_entries format=duration -of json \"$VIDEO_PATH\" > \"$WORK_DIR\u002Fmetadata.json\"\n# audio\nffmpeg -y -i \"$VIDEO_PATH\" -vn -acodec libmp3lame -q:a 2 \"$WORK_DIR\u002Faudio.mp3\"\n```\n\nOutputs: `metadata.json` (read `width`\u002F`height`\u002F`duration`; fps = the `r_frame_rate`\nfraction evaluated, e.g. `30000\u002F1001 → 29.97`) + `audio.mp3`.\n\n### 4. Transcribe\n\n```bash\nnpx hyperframes transcribe \"$WORK_DIR\u002Faudio.mp3\" -d \"$WORK_DIR\" --json --model small.en\n```\n\nLocal **Whisper** — no API key, no proxy, no rate limit. Writes a word-level\n`transcript.json` into the work dir (word `text` + `start` \u002F `end` timestamps).\nRead it for the word \u002F sentence timings that drive card timing in Step 6; group\nwords into sentences yourself at punctuation \u002F pauses if you need segment-level\nchunks.\n\n**Clamp to media duration.** Whisper can return the final word's `end` a hair past the\nactual clip length — clamp every card `endSec` and `composition.durationSeconds` to the\n`metadata.json` duration, or the render will show a black tail past the video.\n\n### 5. Correct Transcript\n\n`transcript.json` is a **flat array of word objects** — `[{ \"text\": \"...\", \"start\": s, \"end\": s }, …]` (no `segments` array, no `words` wrapper; the per-word key is **`text`**). Read it and fix obvious ASR errors:\n\n- Homophones, product names, technical terms, punctuation\n- Edit a word's `text` in place; **preserve its `start` \u002F `end`** timestamps\n- There is no pre-grouped `segments` array — **group words into sentences yourself** (split at terminal punctuation \u002F pauses) when you need segment-level chunks for card timing\n\n### 6. Draft a Lightweight Storyboard (in chat)\n\n**No CLI involved.** Read `transcript.json` + `metadata.json` and design\ncards directly. `storyboard.json` is an agent-internal planning artifact\n— no CLI command consumes it; it exists so you can think clearly\nabout timing and content before writing each card's HTML. Keep the\nshape consistent with the example below so the same outline can drive\nthe composition you author in Step 9:\n\n```json\n{\n  \"schemaVersion\": 3,\n  \"composition\": {\n    \"fps\": 30,\n    \"width\": 1080,\n    \"height\": 1920,\n    \"durationSeconds\": 121.2,\n    \"layout\": \"portrait\",\n    \"themeId\": \"noir\",\n    \"seed\": 42\n  },\n  \"videoTrack\": {\n    \"sourcePath\": \"input-video.mp4\",\n    \"startSec\": 0,\n    \"endSec\": 121.2,\n    \"bounds\": { \"x\": 0, \"y\": 0, \"width\": 1080, \"height\": 1920 }\n  },\n  \"subtitles\": { \"enabled\": false },\n  \"cards\": [\n    {\n      \"id\": \"card-01\",\n      \"intent\": \"Hook with the speaker's anxious midnight question\",\n      \"startSec\": 0.5,\n      \"endSec\": 13.0,\n      \"accentIndex\": 0,\n      \"zone\": \"fullscreen\",\n      \"contentHints\": {\n        \"kicker\": \"AN HONEST QUESTION\",\n        \"title\": \"The soul-searching question at 11 PM\",\n        \"detail\": \"Client's 60-second voice message: 'If the RMB appreciates, does that mean my USD policy is a terrible loss?'\"\n      }\n    }\n  ]\n}\n```\n\n**Required Card fields:**\n\n| field                   | type                                       | purpose                                                                                               |\n| ----------------------- | ------------------------------------------ | ----------------------------------------------------------------------------------------------------- |\n| `id`                    | string                                     | stable id used in card HTML & GSAP selectors                                                          |\n| `intent`                | string                                     | natural-language description; fed to card synthesis                                                   |\n| `startSec` \u002F `endSec`   | number                                     | times in seconds (endSec > startSec)                                                                  |\n| `accentIndex`           | 0 \\| 1 \\| 2 \\| 3 \\| 4                      | which of the 5 theme accent colors this card pulls                                                    |\n| `zone`                  | enum (see below)                           | where on the canvas the card lives                                                                    |\n| `contentHints`          | object                                     | free-form bag; agent puts kicker\u002Ftitle\u002Fdetail\u002Fdata\u002Fquote here                                         |\n| `archetype` (optional)  | string                                     | free-form label you may attach to remember a card's pattern; absent = free-form, which is the default |\n| `transition` (optional) | enum: `cut` \\| `fade` \\| `slide` \\| `wipe` | declarative card-to-card transition                                                                   |\n\n**Five `zone` values:**\n\n| zone              | resolved bounds                                | when to use                             |\n| ----------------- | ---------------------------------------------- | --------------------------------------- |\n| `fullscreen`      | covers whole canvas                            | hero moments, big numbers, mantras      |\n| `whiteboard-area` | inset 40px margin (or 45% of portrait height)  | dense data \u002F annotated content          |\n| `lower-third`     | bottom 30% band                                | annotation over visible video           |\n| `side-panel`      | right 42% (landscape) or bottom 40% (portrait) | data side, video other side             |\n| `video-overlay`   | full canvas, expects mostly-transparent card   | annotation overlays on full-bleed video |\n\nWhen you assemble the composition in Step 9, resolve each card's `zone`\ninto pixel bounds on the card-host wrapper following the table above.\nVideo bounds are set **once** at composition level (`videoTrack.bounds`);\nto make video appear to \"move between cards\", author GSAP tweens against\n`#video-wrap` in the composition's `\u003Cscript>` (see Step 9).\n\n**No prescribed card roles, no prescribed narrative arc.** Cards emerge\nfrom what the video actually says — could be all quotes or all data,\ncould open with a number or with a story. Let the transcript drive the\nrhythm.\n\n**How many takeaways? — auto-infer from duration + density.** No fixed\nupper limit. Pick a **base pace** from the video duration, then adjust\nby **information density**. Only **floor is fixed: minimum 5 cards** so\neven short videos have rhythm.\n\n**Step 1 — base pace by duration** (the natural sec\u002Fcard for medium density):\n\n| video duration     | base pace (sec per card) | rationale                                   |\n| ------------------ | ------------------------ | ------------------------------------------- |\n| \u003C 60s (short reel) | **6–8s**                 | viewers expect fast cuts in short-form      |\n| 60s – 3 min        | **8–12s**                | normal social pace                          |\n| 3 – 10 min         | **12–20s**               | give breathing room; each card carries more |\n| 10 – 30 min        | **20–35s**               | long-form lecture \u002F interview rhythm        |\n| > 30 min           | **30–60s**               | episodic, near-chapter feel                 |\n\n**Step 2 — density multiplier** (multiplies the base pace):\n\n| signal in the transcript                                                                                                    | multiplier | effect                   |\n| --------------------------------------------------------------------------------------------------------------------------- | ---------- | ------------------------ |\n| **High density** — many numbers, distinct claims, staccato pacing, list-like enumeration, every 1–2 sentences is a new idea | **× 0.7**  | cuts faster, more cards  |\n| **Medium density** — mixed flow with both data and narrative                                                                | **× 1.0**  | base pace                |\n| **Low density** — one extended story, repeated reframing, slow reflective pacing, single argument unfolding                 | **× 1.5**  | cuts slower, fewer cards |\n\n**Step 3 — compute:**\n\n```\nsecPerCard = basePace × densityMultiplier\ncardCount  = max(5, round(videoDurationSec \u002F secPerCard))\n```\n\nExamples (notice — **no upper clamp**; long videos naturally produce more cards):\n\n- **30s reel, single punchline (low density)** → 7 × 1.5 = 10.5s\u002Fcard → round(30\u002F10.5)=3 → floor to **5** cards\n- **60s reflective monologue (low density)** → 10 × 1.5 = 15s\u002Fcard → **4** → floor to **5** cards\n- **121s talking-head with rich data (high density)** → 10 × 0.7 = 7s\u002Fcard → **17** cards\n- **5 min interview, mixed density** → 16 × 1.0 = 16s\u002Fcard → **19** cards\n- **10 min deep-dive, high density** → 16 × 0.7 = 11s\u002Fcard → **55** cards\n- **30 min lecture, medium density** → 28 × 1.0 = 28s\u002Fcard → **64** cards\n- **1 hr podcast, low density** → 45 × 1.5 = 67.5s\u002Fcard → **53** cards\n\nWhen a card holds longer than ~15s, plan for a richer card (data block,\nmulti-step reveal, several sub-points unfolding with staggered\nanimations) — a static one-liner gets boring past 8s. For long pieces\nwhere many cards exceed 30s, consider **chunking the timeline into\nsub-compositions** (one .html per chapter, mounted with\n`data-composition-src`) so the GSAP timeline per file stays manageable\n— see the `timeline_track_too_dense` HyperFrames lint warning.\n\n`content` can be a plain string (\"Title: annualized 5.69%\\nNotes: ...\") or any JSON\nshape that captures the data. The agent decides the shape per card.\n\n**Optional outro.** This skill ships **no fixed brand outro**. If the user wants a closing card, design a neutral one yourself (wordmark + one-line tagline, ~1.5-2s, fade in -> short hold -> fade out), append it to `cards[]`, and extend `composition.durationSeconds` to its `endSec`. Otherwise end on the last content card.\n\n### 7. Decide Render Strategy\n\n#### Confirm Visual Direction with User (DO THIS FIRST)\n\nBefore you start designing cards or deciding bounds, **ask the user to\npick the output ratio, the layout, the style, and the card-density\npreset**. Frames are auto-selected from the chosen layout × style\ncombination (see \"Auto-pick frame\" table below). Before sending the\nquestion, **precompute two things**:\n\n1. **`recommendedRatio`** from the source video's aspect ratio\n   (`metadata.json` width \u002F height):\n   - `sourceAspect = width \u002F height`\n   - `sourceAspect ≥ 1.5` (≥ ~3:2 wide) → recommend **`16:9`**\n   - `sourceAspect ≤ 0.7` (≤ ~9:13 tall) → recommend **`9:16`**\n   - `0.7 \u003C sourceAspect \u003C 1.5` (near-square) → recommend **`4:5`**\n\n   Mark the recommended option's label with \" (recommended · matches source video X:Y)\"\n   so the user sees why it's recommended.\n\n2. **`autoCount`** from Step 6 (`max(5, round(videoSec \u002F (basePace ×\ndensityMultiplier)))`) so the \"auto\" option's label can show the\n   concrete number.\n\n**Environment compatibility — pick the best available question channel.**\nNot every runtime exposes the same structured-question tool. Apply this\norder:\n\n1. **Native clarification tool** — use the structured 4-question call below.\n2. **Other native clarification tool** (e.g. `ask_question`,\n   `request_user_input`, IDE-specific prompt) — use that tool with the\n   same 4 question texts and option lists. Preserve the recommendation\n   markers and the precomputed values.\n3. **No native tool** (Codex CLI, plain text-only runtimes) — **ask\n   directly in normal conversation**. Use the plain-text template at the\n   end of this section. Keep it to **one message, 4 numbered questions**\n   (the global cap is 2–5 questions per round; we stay inside it).\n\nRules that apply to every channel:\n\n- Ask **at most 2–5 questions per round**. Our 4 here fits.\n- Even if missing info doesn't block rendering, **ask once to confirm\n  the parameters that materially affect the final output** (ratio,\n  layout, style, cardCount).\n- If the user has already pre-approved defaults (\"just use defaults\",\n  \"no need to ask\", \"auto-pick everything\"), asked you not to ask, or the\n  run carries an ongoing autonomous signal (\"surprise me\" \u002F \"decide for me\" —\n  `..\u002Fhyperframes-core\u002Freferences\u002Fbrief-contract.md` § 1) — **skip\n  the question entirely** and use: `recommendedRatio`, `layout=\"stack\"`\n  (safest cross-ratio default), `style` chosen from transcript tone in\n  the most neutral group (editorial\u002Fdata), `autoCount`. Tell the user\n  what you picked in one sentence and continue.\n\n**Channel A — native `AskUserQuestion`:**\n\n```\n\u002F\u002F Precompute before the call:\n\u002F\u002F   recommendedRatio = \"16:9\" | \"9:16\" | \"4:5\"\n\u002F\u002F   autoCount        = integer (from Step 6)\n\nAskUserQuestion({\n  questions: [\n    {\n      question: \"Output video aspect ratio (canvas):\",\n      header: \"Aspect ratio\",\n      multiSelect: false,\n      \u002F\u002F Reorder so the recommended option appears FIRST (per AskUserQuestion convention).\n      \u002F\u002F Append \" (recommended · matches source video W×H)\" to the recommended option's label.\n      options: [\n        { label: \"16:9 (1920×1080) landscape\", description: \"TV \u002F YouTube \u002F desktop playback. Most natural when the source video is already landscape; widest canvas.\" },\n        { label: \"9:16 (1080×1920) portrait\", description: \"TikTok \u002F Reels \u002F short-form mobile. Most natural for portrait source; native mobile experience.\" },\n        { label: \"4:5 (1080×1350) near-portrait\", description: \"Instagram feed \u002F WeChat Moments. Best when source is near-square or you want to cover both platforms.\" }\n      ]\n    },\n    {\n      question: \"Choose the overall layout: how should the video and cards coexist on the canvas?\",\n      header: \"Layout\",\n      multiSelect: false,\n      options: [\n        { label: \"side-by-side (split)\",  description: \"Video and card each take half the canvas. Most stable for interview \u002F data side-by-side; clear visual separation.\" },\n        { label: \"top-bottom (stack)\",    description: \"Video on top (~52%), card below. Classic combo of speaker face + summary card; works well in portrait too.\" },\n        { label: \"picture-in-picture (pip)\", description: \"Card fills the canvas, video shrinks to a rounded corner window. Use when content is primary and speaker is secondary.\" },\n        { label: \"full-screen overlay (overlay)\", description: \"Video plays full-bleed, card floats as a glass layer on top. Strong cinematic \u002F emotional feel.\" }\n      ]\n    },\n    {\n      question: \"Choose the card visual style (style):\",\n      header: \"Style group\",\n      multiSelect: false,\n      \u002F\u002F NOTE: these 3 groups intentionally match the frame auto-pick matrix\n      \u002F\u002F rows below, so picking a group resolves both `style` group AND the\n      \u002F\u002F frame matrix column in one step. Memberships are mutually exclusive.\n      options: [\n        { label: \"warm paper (warm-paper)\", description: \"academic notebook · editorial big-type · whiteboard hand-drawn · xhs social. Best for interview reflections, product launches, lifestyle, emotional stories.\" },\n        { label: \"clinical \u002F cold (clinical)\",   description: \"audit magazine · swiss grid · terminal CLI · minimal modern. Best for financial analysis, investigative reports, technical tutorials, serious presentations.\" },\n        { label: \"experimental \u002F avant-garde (experimental)\", description: \"geom color-clash geometry · spotlight dark-background. Best for short-form highlights, product launches, strong emotion, cinematic feel.\" }\n      ]\n    },\n    {\n      question: \"Card count (takeaway pacing): how many cards to cut?\",\n      header: \"Card count\",\n      multiSelect: false,\n      options: [\n        { label: \"Auto (recommended) · approx N cards\", description: \"Inferred automatically from video duration and information density (see Step 6 rules). This run estimates approx N cards. Substitute the real N (your autoCount) into the label.\" },\n        { label: \"Fewer · approx round(N × 0.6) cards\", description: \"Sparser cuts, each card holds longer — suits reflective \u002F slow-paced content.\" },\n        { label: \"More · approx round(N × 1.5) cards\", description: \"Tighter cuts, faster rhythm — suits staccato \u002F data-dense \u002F short-form highlight content.\" }\n      ]\n    }\n  ]\n})\n```\n\n**About \"Other\"** — `AskUserQuestion` automatically adds an \"Other\" option to the card count question. The user can type a number directly (e.g. \"8\", \"20\") as the cardCount target. Parse the input as an integer: if parsing succeeds → use that value (minimum 5 as a floor); if parsing fails → fall back to \"auto\".\n\n**Channel B — plain-text fallback** (Codex CLI, runtimes without a\nnative question tool). Post this as one normal message, then wait for\nthe reply. Bullet-style 1\u002F2\u002F3\u002F4 keeps the reply parseable:\n\n```\nI need to confirm four visual decisions with you before I start cutting cards:\n\n1) Output aspect ratio (canvas):\n   A. 16:9 landscape (1920×1080) — TV \u002F YouTube \u002F desktop playback\n   B. 9:16 portrait (1080×1920) — TikTok \u002F Reels \u002F short-form mobile\n   C. 4:5 near-portrait (1080×1350) — Instagram feed \u002F works for both platforms\n   ▸ My recommendation:  \u003CrecommendedRatio>  (matches source video W×H = \u003CsourceW>×\u003CsourceH>)\n\n2) Overall layout (how video & card coexist):\n   A. split   side-by-side (50\u002F50)\n   B. stack   top-bottom (video top, card bottom)\n   C. pip     picture-in-picture (card full canvas, video rounded corner window)\n   D. overlay full-screen glass overlay (video full-bleed, card glass layer)\n\n3) Card style group (maps to frame auto-pick matrix, pick 1 of 3):\n   A. warm paper (warm-paper)      (academic \u002F editorial \u002F whiteboard \u002F xhs)\n   B. clinical \u002F cold (clinical)   (audit \u002F swiss \u002F terminal \u002F minimal)\n   C. experimental (experimental)  (geom \u002F spotlight)\n\n4) Card count (takeaway pacing):\n   A. Auto (recommended) — approx \u003CautoCount> cards\n   B. Fewer — approx round(\u003CautoCount> × 0.6) cards\n   C. More — approx round(\u003CautoCount> × 1.5) cards\n   D. Give me a specific number (e.g. \"8\", \"20\")\n\nReply format: \"1A 2C 3B 4A\" or natural language is fine.\nIf you want all recommended defaults, reply \"default\" \u002F \"auto\" \u002F \"use all recommendations\".\n```\n\nParsing the plain-text reply:\n\n- Accept loose formats: `\"1A 2C 3B 4A\"`, `\"A C B A\"`, `\"16:9 \u002F pip \u002F\ndata \u002F auto\"`, full sentences, or `default`.\n- If any answer is ambiguous → re-ask only the ambiguous ones (still\n  inside the 2–5 cap).\n- If the user says \"default \u002F auto \u002F use all recommendations\" → skip without re-asking.\n\nAfter the user answers (any channel):\n\n1. **Resolve the output canvas** from the ratio answer — these are the\n   exact `storyboard.composition.width \u002F height` values to write:\n\n   | user choice | composition.width × height | storyboard.layout field                                       |\n   | ----------- | -------------------------- | ------------------------------------------------------------- |\n   | `16:9`      | **1920 × 1080**            | `\"landscape\"`                                                 |\n   | `9:16`      | **1080 × 1920**            | `\"portrait\"`                                                  |\n   | `4:5`       | **1080 × 1350**            | `\"portrait\"` (schema treats 4:5 as portrait — height > width) |\n\n   For **4:5 bounds inside `references\u002Flayouts\u002F*.html`** — those files\n   only document landscape (1920×1080) and portrait (1080×1920). For\n   4:5 (1080×1350) derive bounds by **proportional scaling from\n   portrait**: keep horizontal values, scale vertical values by\n   `1350\u002F1920 ≈ 0.703`. Example: `overlay` portrait card =\n   `{ x: 24, y: 1280, w: 1032, h: 564 }` → 4:5 card =\n   `{ x: 24, y: round(1280 × 0.703), w: 1032, h: round(564 × 0.703) }`\n   = `{ x: 24, y: 900, w: 1032, h: 397 }`.\n\n2. **Map the style group to a specific style** by looking at the\n   transcript tone — pick the one that best fits, but stay inside the\n   user's chosen group. If you're unsure between two specific styles\n   inside the group, send a second `AskUserQuestion` with those 2–4\n   specific style options.\n\n3. **Resolve final cardCount** from the density answer:\n\n   | user choice             | final cardCount                           |\n   | ----------------------- | ----------------------------------------- |\n   | Auto (recommended)      | the `autoCount` you already computed      |\n   | Fewer                   | `max(5, round(autoCount × 0.6))`          |\n   | More                    | `round(autoCount × 1.5)` (no upper clamp) |\n   | Other = \"\u003Cn>\" (integer) | `max(5, parseInt(n))`                     |\n   | Other = anything else   | fall back to `autoCount`                  |\n\n4. **Auto-pick the video frame** from this table (frames don't ask the\n   user — they follow from layout × style):\n\n   | layout    | warm-paper styles (academic \u002F whiteboard \u002F editorial \u002F xhs) | clinical styles (audit \u002F swiss \u002F terminal \u002F minimal) | experimental styles (geom \u002F spotlight) |\n   | --------- | ----------------------------------------------------------- | ---------------------------------------------------- | -------------------------------------- |\n   | `split`   | `polaroid`                                                  | `hairline`                                           | `clean`                                |\n   | `stack`   | `polaroid`                                                  | `hairline`                                           | `clean`                                |\n   | `pip`     | `clean` (pip pill already has chrome)                       | `clean`                                              | `clean`                                |\n   | `overlay` | `clean` (full-bleed forbids deco frames)                    | `clean`                                              | `clean`                                |\n\n5. **Tell the user what you chose** in one sentence — ratio (+ canvas\n   size), layout, specific style, frame, and final cardCount — then\n   proceed with the rest of Step 7 (per-card layouts, motion patterns).\n6. Record the five values (ratio \u002F layout \u002F style \u002F frame \u002F cardCount)\n   in working memory (no schema field needed); you'll reference them\n   while writing each card's HTML in Step 8 and while reading the\n   matching `references\u002F\u003Cdim>\u002F\u003Ckey>.html` for tokens and structure.\n\nIf the user picks an answer via \"Other\" with a free-text style name not\nin the 10-style library, treat it as a hint to design a fresh card\nvisual yourself, but still anchor on the chosen layout's bounds.\n\n#### Render Strategy Inputs\n\nWith ratio \u002F layout \u002F style \u002F cardCount \u002F frame locked from Step 7.0,\nthe remaining per-card decisions are:\n\n- **Source-video fit inside the GSAP target**: video element has\n  `object-fit: cover` and is clipped to `#video-wrap`'s tween bounds.\n  If you want NO cropping (e.g. portrait source on landscape canvas\n  shouldn't get its top\u002Fbottom chopped), aim the tween at a rect that\n  matches the source's aspect ratio and let surrounding canvas show\n  through (or fill with the card \u002F a backdrop).\n- **`card.zone` per card**: derive from your chosen composition layout\n  (split → side-panel, stack → lower-third, pip → fullscreen, overlay\n  → video-overlay), OR pick a different zone for one-off variants\n  (fullscreen for hero \u002F quote, whiteboard-area for dense data).\n- **`accentIndex` per card**: each card pulls one of the 5 theme accent\n  colors. Vary across cards for rhythm; reuse the same index when two\n  cards belong to the same narrative beat.\n- **Motion vocabulary**: pick 2–3 repeatable patterns from\n  `data-anim` kinds (see the table later) and stick to them so the\n  composition feels coherent.\n\nPick from these `themeId` palettes (use them as `--accent-N` \u002F\n`--bg` \u002F `--text` CSS variables in your composition `\u003Cstyle>` block):\n\n| themeId | accent palette (5 colors)                 | board bg          | text      |\n| ------- | ----------------------------------------- | ----------------- | --------- |\n| classic | `#1971c2 #e03131 #2f9e44 #e8590c #9c36b5` | `#FFF9E3` (paper) | `#1e1e1e` |\n| noir    | `#4cc9f0 #f72585 #4ade80 #fb923c #a78bfa` | `#1a1a1a`         | `#f1f1f1` |\n| mint    | `#0077b6 #d62828 #2d6a4f #e76f51 #7209b7` | `#e8faf0`         | `#1b4332` |\n| craft   | `#bf5700 #d62728 #6c757d #e9b54a #3d5a80` | `#f6efe1`         | `#2d2d2d` |\n| slate   | `#0ea5e9 #ef4444 #22c55e #f97316 #a855f7` | `#1e293b`         | `#f1f5f9` |\n| mono    | `#000 #555 #888 #aaa #ccc`                | `#fff`            | `#000`    |\n\nAvailable fonts (woff2 in `\u003CSKILL_DIR>\u002Fassets\u002Ffonts\u002F`, staged to work dir in Step 9): `Caveat` (handwriting),\n`LXGW WenKai TC` (Chinese hand-script), `Inter` (modern sans), `Virgil`\n(geometric hand). Reference via `@font-face` or `font-family` directly.\n\nFor inspiration on visual patterns, `\u003CSKILL_DIR>\u002Freferences\u002Fstyles\u002F`\nships 10 self-contained reference cards (academic \u002F editorial \u002F minimal\n\u002F spotlight \u002F geom \u002F whiteboard \u002F audit \u002F terminal \u002F swiss \u002F xhs) that\nyou can copy as starting points — but **do not feel constrained to\nmatch any of these**. Each card is your own design.\n\n#### Visual Design Library (\u003CSKILL_DIR>\u002Freferences\u002F)\n\nBeyond the composition-level `themeId`, the skill ships a richer **reference\nlibrary** at `\u003CSKILL_DIR>\u002Freferences\u002F` covering three **orthogonal**\nvisual dimensions you can freely mix:\n\n```\nStyle  ×  Layout  ×  VideoFrame\n (10)      (4)         (3)\n```\n\n| dimension  | keys                                                                                              | what it decides                                                          |\n| ---------- | ------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------ |\n| **style**  | `academic` `editorial` `minimal` `spotlight` `geom` `whiteboard` `audit` `terminal` `swiss` `xhs` | the card's visual language — fonts, colors, ornament, layout-within-card |\n| **layout** | `split` `stack` `pip` `overlay`                                                                   | how the source video and the card share the canvas                       |\n| **frame**  | `clean` `hairline` `polaroid`                                                                     | the decorative chrome around the video element                           |\n\nRead `\u003CSKILL_DIR>\u002Freferences\u002FDESIGN_INDEX.md`\nfor the full matrix and a loose decision guide (interview \u002F product launch \u002F data analysis \u002F\nsocial clip \u002F technical tutorial \u002F emotional story …). When you decide to use a specific\nstyle \u002F layout \u002F frame, Read the corresponding file:\n\n- `references\u002Fstyles\u002F\u003Ckey>.html` — self-contained card fragment with that\n  style's CSS tokens (colors, fonts, padding, ornament) and a placeholder\n  takeaway. Copy the `.card[data-card-id=\"ref-\u003Ckey>\"]` style block, rename\n  the data-card-id to your card's id, swap the placeholder content for the\n  real takeaway, and you're done.\n- `references\u002Flayouts\u002F\u003Ckey>.html` — exact `videoBounds` + `cardBounds` for\n  both landscape and portrait, with a copy-paste JSON snippet for\n  `storyboard.json`'s per-card `layout` field.\n- `references\u002Fframes\u002F\u003Ckey>.html` — decorative HTML to add as a sibling of\n  `#video-wrap`, plus placement instructions for the composition CSS.\n\nPick `style × layout × frame` **per card** — you can change all three\nbetween cards as long as the transitions read smoothly. A common rhythm:\nopen `editorial × overlay × clean`, switch to `audit × split × hairline`\nfor the data card, close on `whiteboard × pip × polaroid`.\n\nThe 10 styles are skill-side design tokens, **not composition-level themes** —\nthey don't need to be declared in `storyboard.composition`; they live\ninside each card's HTML. The `themeId` field can still pick a\ncomposition-level palette (table above) that controls page-body background\nand video border chrome.\n\n#### Layout Compositions (Card + Video)\n\nTwo coordinated decisions per card define how it shares the canvas with\nthe source video:\n\n- **`card.zone`** (declared in `storyboard.json`) — one of the 5 schema\n  values; resolve it into pixel bounds (per the table in Step 6) when\n  you write the card-host wrapper's inline `style` in Step 9.\n- **`#video-wrap` bounds at this card's time window** (declared\n  imperatively in the composition's GSAP timeline) — the agent tweens\n  `#video-wrap` to a target rect for each layout transition.\n\nSchema does NOT store per-card video bounds. `videoTrack.bounds` is\n**one-time** at composition level (defaults to full canvas). Video\n\"moving\" between cards is purely a GSAP animation authored in\n`index.html`. There is no `card.layout` field — earlier versions of this\ndoc invented one; the real schema only has `card.zone`.\n\n**4 composition layouts** (from `references\u002Flayouts\u002F`) — each is a\nrecipe pairing a `zone` with a `#video-wrap` tween target:\n\n| composition layout | recommended `card.zone` | GSAP target for `#video-wrap` (landscape 1920×1080)                       | GSAP target for `#video-wrap` (portrait 1080×1920)                | when to use                                     |\n| ------------------ | ----------------------- | ------------------------------------------------------------------------- | ----------------------------------------------------------------- | ----------------------------------------------- |\n| `split`            | `side-panel`            | `{ left: 960, top: 0, width: 960, height: 1080 }`                         | `{ left: 0, top: 960, width: 1080, height: 960 }` (bottom half)   | speaker + data side-by-side \u002F 50:50 weight      |\n| `stack`            | `lower-third`           | `{ left: 14, top: 14, width: 1892, height: 548 }` (top 52%)               | `{ left: 0, top: 0, width: 1080, height: 844 }` (top 44%)         | speaker on top + summary card below             |\n| `pip`              | `fullscreen`            | `{ left: 1480, top: 760, width: 400, height: 300 }` + add `.framed` class | `{ left: 690, top: 28, width: 360, height: 203 }` + add `.framed` | content-heavy card + corner pip                 |\n| `overlay`          | `video-overlay`         | `{ left: 0, top: 0, width: 1920, height: 1080 }` (full-bleed)             | `{ left: 0, top: 0, width: 1080, height: 1920 }`                  | cinematic \u002F dramatic \u002F glass card on full video |\n\nFor 4:5 (1080×1350), scale portrait y\u002Fh values by `1350\u002F1920 ≈ 0.703`\n(see Step 7.0 Channel A \u002F Channel B `recommendedRatio` resolution\ntable).\n\n**Other zone values for one-off variants** (still uses `card.zone`; no\nfake \"layout\" field):\n\n| `zone`            | resolved bounds                                        | common use                            |\n| ----------------- | ------------------------------------------------------ | ------------------------------------- |\n| `fullscreen`      | covers whole canvas                                    | hero card, video tweens to hidden\u002Fpip |\n| `whiteboard-area` | inset 40px margin (landscape) or bottom 45% (portrait) | dense data card, free margins         |\n| `lower-third`     | bottom 30% band                                        | talking-head annotation               |\n| `side-panel`      | right 42% (landscape) or bottom 40% (portrait)         | sidebar \u002F \"split\" recipe              |\n| `video-overlay`   | full canvas; expect transparent card root              | glass overlay on full-bleed video     |\n\nYou can mix recipes per card — choose `card.zone` based on what suits\nthe moment, then write the GSAP tween for `#video-wrap` between cards.\n\n#### Storyboard Render Contract\n\n`storyboard.json` is an agent-internal planning artifact — no CLI\ncommand parses it. It exists to keep your timing and content decisions\nexplicit before you write each card's HTML. Stick to the v3-style\nshape below so the same outline drives the composition you assemble in\nStep 9.\n\nRequired structure (see Step 6 for the full example):\n\n- `schemaVersion: 3`\n- `composition: { fps, width, height, durationSeconds, layout, themeId, seed }` — note `durationSeconds`\u002F`fps`\u002F`themeId`\u002F`layout` live **inside** `composition`, NOT at top level\n- `videoTrack: { sourcePath, startSec, endSec, bounds? }` — video bounds default to full canvas\n- `subtitles: { enabled, ... }`\n- `cards[]` — each card has the 6 required fields: `id`, `intent`, `startSec`, `endSec`, `accentIndex`, `zone`, `contentHints`\n\nRules:\n\n- Card times stay inside `composition.durationSeconds` and should not overlap unless intentional (use `data-track-index` to control z-order when they do).\n- Visual details live in card HTML fragments (Step 8), NOT in `contentHints`. `contentHints` is your own structured prompt for designing the card; the rendered look is the HTML.\n- Keep the storyboard shape stable — even though nothing parses it, you read it back while authoring Step 8\u002F9, and consistency keeps card IDs and timing in sync.\n- Agent-side decisions like \"I picked overlay × geom × clean\" do NOT belong in `storyboard.json` — keep them in working memory and use them when authoring card HTML + GSAP tweens.\n\n**Transparent card backgrounds for cards that share canvas with video.**\nWhen the GSAP tween leaves video visible behind\u002Fbeside the card (overlay\nrecipe, pip recipe, or any `card.zone = 'lower-third' | 'video-overlay'`\nmoment), the card's `.root` MUST NOT paint a full opaque background —\notherwise it occludes the video. Two patterns:\n\n```css\n\u002F* Pattern A: transparent root, page body provides the cream backdrop *\u002F\nhtml,\nbody {\n  background: var(--bg);\n}\n.card[data-card-id=\"card-X\"] .root {\n  background: transparent;\n}\n\n\u002F* Pattern B: explicit per-card background ONLY for fullscreen cards *\u002F\n.card[data-card-id=\"card-hero\"] .root {\n  background: var(--bg);\n}\n.card[data-card-id=\"card-overlay\"] .root {\n  background: transparent;\n}\n```\n\nFor `side-panel`-zone cards (split recipe), the card-host is already\nonly half the canvas, so an opaque card bg is fine — it only covers its\nhalf.\n\n### 8. Write Each Card's HTML\n\nCreate `$WORK_DIR\u002Fpublic\u002Fcards\u002F{card-id}.html` for each card. Each file\ncontains a single rooted HTML fragment that follows this contract:\n\n#### Card HTML Contract\n\n```html\n\u003Cdiv class=\"card\" data-card-id=\"{cardId}\">\n  \u003Cstyle>\n    \u002F* MUST: every rule starts with .card[data-card-id=\"{cardId}\"] *\u002F\n    .card[data-card-id=\"card-01\"] .root {\n      width: 100%; height: 100%;\n      display: flex; ...;\n      font-family: 'Caveat', 'LXGW WenKai TC', serif;\n      color: var(--text);\n      background: var(--bg);\n    }\n    .card[data-card-id=\"card-01\"] .title { font-size: 84px; ... }\n  \u003C\u002Fstyle>\n\n  \u003Cdiv class=\"root\">\n    \u003Ch1\n      id=\"card-01-title\"\n      data-anim=\"kinetic-chars\"\n      data-anim-at=\"0.3\"\n      data-anim-duration=\"0.5\"\n      data-anim-stagger=\"0.04\"\n      data-anim-pattern=\"pop\"\n    >\n      \u003Cspan class=\"char\">S\u003C\u002Fspan>\n      \u003Cspan class=\"char\">u\u003C\u002Fspan>\n    \u003C\u002Fh1>\n    \u003Cdiv\n      id=\"card-01-line\"\n      data-anim=\"grow-x\"\n      data-anim-at=\"0.65\"\n      data-anim-duration=\"0.5\"\n      data-anim-target-w=\"420\"\n      style=\"width:0;height:8px;background:var(--accent-0);border-radius:4px;\"\n    >\u003C\u002Fdiv>\n  \u003C\u002Fdiv>\n\u003C\u002Fdiv>\n```\n\n**Hard rules** (`hyperframes` lint will reject violations):\n\n- Single root `\u003Cdiv class=\"card\" data-card-id=\"{cardId}\">`\n- Inline `\u003Cstyle>` rules MUST be prefixed with the scope selector above\n- **No `\u003Cscript>` tags**\n- **No external URLs** in `src=` \u002F `href=` (no CDN, no remote fonts)\n- **No inline event handlers** (`onclick=` etc.)\n- All assets via relative paths into the same `public\u002F` directory\n- Colors via `var(--accent-N)` etc. for portability across themes\n\n**Animations are declared, not coded.** Use `data-anim-*` attributes\nonly; never write `\u003Cscript>` to animate. You compile every `data-anim-*`\ndeclaration into the single master GSAP timeline in Step 9.\n\n#### Card Sizing — Mobile-First in Portrait\n\nThe 10 `references\u002Fstyles\u002F*.html` are sized for a **1920×1080 landscape**\npreview. When `storyboard.layout = \"portrait\"` (1080×1920, the dominant\ncase for social \u002F mobile), **scale every visual size up** — phones hold\nthe screen close, and the same pixel count reads smaller than on a\nlandscape TV-style canvas.\n\n| token                     | landscape baseline | **portrait target** | scale         |\n| ------------------------- | ------------------ | ------------------- | ------------- |\n| title (h1\u002Fh2 hero)        | 64–96px            | **88–132px**        | ×1.35         |\n| detail \u002F body             | 24–30px            | **30–40px**         | ×1.30         |\n| kicker \u002F chip label       | 14–16px            | **18–22px**         | ×1.30         |\n| timecode \u002F meta           | 12–14px            | **16–18px**         | ×1.30         |\n| data block primary number | 48–60px            | **64–88px**         | ×1.40         |\n| line-height multiplier    | 1.05–1.5           | same                | (don't scale) |\n\n**Rule of thumb:** `portraitPx = round(landscapePx × 1.3)`, then floor\nto a nearby 4px multiple for visual rhythm. Hero headlines may go up to\n×1.4; small meta text stays at ×1.2 to avoid crowding.\n\nPadding **shrinks slightly** in portrait — the card is narrower so big\nlandscape padding (40–64px) eats too much width. Use 24–36px horizontal\npadding in portrait.\n\nIf you're producing a single card that must work in **both** layouts,\nprefer a `@container` query on the card root over hard-coding sizes:\n\n```css\n.card[data-card-id=\"X\"] .root {\n  container-type: inline-size;\n}\n.card[data-card-id=\"X\"] .title {\n  font-size: clamp(64px, 8.5cqi, 132px);\n}\n.card[data-card-id=\"X\"] .detail {\n  font-size: clamp(24px, 3.2cqi, 40px);\n}\n```\n\nBut for most cards, a single layout choice is fine — just pick the size\ntable column that matches the storyboard's `layout` field.\n\n#### Available `data-anim` Kinds\n\n| kind            | use for             | key params                                                                                      |\n| --------------- | ------------------- | ----------------------------------------------------------------------------------------------- |\n| `fade-in`       | enter               | `at`, `duration`, `ease?`                                                                       |\n| `fade-out`      | exit                | `at`, `duration`, `ease?`                                                                       |\n| `slide-in`      | slide enter         | `at`, `duration`, `from=left\\|right\\|top\\|bottom`, `distance`                                   |\n| `kinetic-chars` | per-char pop        | `at`, `duration`, `stagger`, `pattern=pop\\|fade` — element needs `\u003Cspan class=\"char\">` children |\n| `typewriter`    | per-char fade       | same as kinetic-chars but slower default stagger                                                |\n| `count-up`      | animate number      | `at`, `duration`, `from`, `to`, `format=.0f\\|.1f\\|.2f\\|,d`                                      |\n| `draw-path`     | SVG path reveal     | `at`, `duration` — element should be a `\u003Cpath>`                                                 |\n| `grow-y`        | bar height          | `at`, `duration`, `target-h` (px) — element starts `height:0`                                   |\n| `grow-x`        | bar width           | `at`, `duration`, `target-w` (px) — element starts `width:0`                                    |\n| `scale-pop`     | pop entrance        | `at`, `duration`                                                                                |\n| `blur-in`       | unfocused → focused | `at`, `duration`                                                                                |\n| `mask-reveal`   | clip reveal         | `at`, `duration`, `direction=left\\|right\\|top\\|bottom`                                          |\n| `morph-to`      | tween any CSS       | `at`, `duration`, `props='{...JSON...}'`                                                        |\n\n`data-anim-at` is **seconds relative to the card's startSec** — when you\ncompile each declaration into the GSAP timeline in Step 9, add the\ncard's `startSec` to get the absolute time and quantize to 1\u002Ffps.\n\n### 9. Assemble the Composition HTML\n\nStage the assets and write `$WORK_DIR\u002Fpublic\u002Findex.html`:\n\n```bash\n# SKILL_DIR is injected by the host (\"Base directory for this skill: …\")\nSKILL_DIR=\"\u003CSKILL_DIR>\"\n\nmkdir -p \"$WORK_DIR\u002Fpublic\u002Ffonts\" \"$WORK_DIR\u002Fpublic\u002Fvendor\" \"$WORK_DIR\u002Fpublic\u002Fcards\"\ncp -n \"$SKILL_DIR\u002Fassets\u002Ffonts\u002F\"*            \"$WORK_DIR\u002Fpublic\u002Ffonts\u002F\"\ncp -n \"$SKILL_DIR\u002Fassets\u002Fvendor\u002Fgsap.min.js\" \"$WORK_DIR\u002Fpublic\u002Fvendor\u002F\"\n# stage the input video — RE-ENCODE with dense keyframes. Sources with a sparse GOP\n# (keyframe interval > ~1s) freeze on seek in the renderer (a frozen frame under the\n# overlays); -g \u002F -keyint_min set to your composition fps make every frame seekable.\n# (Set both to your fps — 30 shown; use 24\u002F25\u002F60 to match.)\nffmpeg -y -i \"$VIDEO_PATH\" -c:v libx264 -crf 18 -g 30 -keyint_min 30 \\\n  -pix_fmt yuv420p -movflags +faststart -c:a aac \"$WORK_DIR\u002Fpublic\u002Finput-video.mp4\"\n```\n\n#### Composition Template\n\n```html\n\u003C!doctype html>\n\u003Chtml lang=\"en\">\n  \u003Chead>\n    \u003Cmeta charset=\"utf-8\" \u002F>\n    \u003Cstyle>\n      @font-face {\n        font-family: \"Caveat\";\n        src: url(\"fonts\u002FCaveat-400-latin.woff2\") format(\"woff2\");\n        font-weight: 400;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"Caveat\";\n        src: url(\"fonts\u002FCaveat-700-latin.woff2\") format(\"woff2\");\n        font-weight: 700;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"LXGW WenKai TC\";\n        src: url(\"fonts\u002FLXGWWenKaiTC-400-latin.woff2\") format(\"woff2\");\n        font-weight: 400;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"Inter\";\n        src: url(\"fonts\u002FInter-400-latin.woff2\") format(\"woff2\");\n        font-weight: 400;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"Inter\";\n        src: url(\"fonts\u002FInter-700-latin.woff2\") format(\"woff2\");\n        font-weight: 700;\n        font-display: block;\n      }\n      @font-face {\n        font-family: \"Virgil\";\n        src: url(\"fonts\u002FVirgil.woff2\") format(\"woff2\");\n        font-display: block;\n      }\n\n      :root {\n        \u002F* Pick from the themeId palette table in Step 7 — example: classic *\u002F\n        --bg: #fff9e3;\n        --text: #1e1e1e;\n        --accent-0: #1971c2;\n        --accent-1: #e03131;\n        --accent-2: #2f9e44;\n        --accent-3: #e8590c;\n        --accent-4: #9c36b5;\n        --font-family: \"Caveat\", \"LXGW WenKai TC\", serif;\n      }\n      * {\n        box-sizing: border-box;\n      }\n      \u002F* Body font-family MUST list concrete font names (not just var(--font-family)) —\n   the HyperFrames renderer's static analyzer doesn't expand CSS variables when\n   resolving fonts, so a var-only chain triggers `font_family_without_font_face`\n   lint and falls back to a generic. Use the concrete chain here; cards that\n   want the theme font can still reference var(--font-family) internally. *\u002F\n      html,\n      body {\n        margin: 0;\n        padding: 0;\n        width: 100%;\n        height: 100%;\n        overflow: hidden;\n        background: #000;\n        font-family: \"Inter\", \"Caveat\", \"LXGW WenKai TC\", ui-sans-serif, system-ui, sans-serif;\n      }\n      #stage {\n        position: relative;\n        width: 100%;\n        height: 100%;\n        overflow: hidden;\n      }\n\n      \u002F* video-wrapper holds the source video. Its position \u002F size are animated\n   over time by the master timeline (one tween per layout transition). *\u002F\n      .video-wrapper {\n        position: absolute;\n        left: 0;\n        top: 0;\n        width: 1920px;\n        height: 1080px;\n        overflow: hidden;\n        border-radius: 0;\n        box-shadow: none;\n      }\n      .video-wrapper video {\n        width: 100%;\n        height: 100%;\n        object-fit: cover;\n      }\n\n      .card-host {\n        position: absolute;\n        pointer-events: none;\n        overflow: hidden;\n      }\n      .card-host .card {\n        position: relative;\n        width: 100%;\n        height: 100%;\n        overflow: hidden;\n      }\n      .card-host .char {\n        display: inline-block;\n        visibility: visible;\n      }\n\n      \u002F* Subtle drop shadow + rounded corners for non-fullscreen video framings *\u002F\n      .video-wrapper.framed {\n        border-radius: 16px;\n        box-shadow: 0 12px 40px rgba(0, 0, 0, 0.35);\n      }\n    \u003C\u002Fstyle>\n  \u003C\u002Fhead>\n  \u003Cbody>\n    \u003Cdiv\n      id=\"stage\"\n      data-composition-id=\"talking-head-recut\"\n      data-start=\"0\"\n      data-duration=\"121.2\"\n      data-fps=\"30\"\n      data-width=\"1920\"\n      data-height=\"1080\"\n    >\n      \u003C!-- Layer 1: source video — initial position matches card-01's layout -->\n      \u003Cdiv class=\"video-wrapper\" id=\"video-wrap\">\n        \u003Cvideo\n          id=\"bg-video\"\n          src=\"input-video.mp4\"\n          muted\n          playsinline\n          data-start=\"0\"\n          data-duration=\"121.2\"\n          data-track-index=\"1\"\n        >\u003C\u002Fvideo>\n      \u003C\u002Fdiv>\n      \u003C!-- Preserve the source program audio while the visual video stays muted. -->\n      \u003Caudio\n        id=\"source-audio\"\n        src=\"input-video.mp4\"\n        data-start=\"0\"\n        data-duration=\"121.2\"\n        data-track-index=\"10\"\n        data-volume=\"1\"\n      >\u003C\u002Faudio>\n\n      \u003C!-- Layer 2: each card-host sits at the bounds dictated by its layout. -->\n      \u003C!-- IMPORTANT: every card-host MUST carry BOTH \"card-host\" and \"clip\" classes. -->\n      \u003C!--   - \"card-host\"  → our positioning + pointer-events styles                 -->\n      \u003C!--   - \"clip\"       → HyperFrames runtime uses this to enforce visibility     -->\n      \u003C!--                    only during data-start … data-start+data-duration.      -->\n      \u003C!--                    Without \"clip\" the host stays visible the whole video   -->\n      \u003C!--                    (lint: timed_element_missing_clip_class).               -->\n      \u003C!-- Example: card-01 with zone=\"fullscreen\" → card-host covers (0,0,1920,1080) -->\n      \u003Cdiv\n        class=\"card-host clip\"\n        data-card-id=\"card-01\"\n        data-start=\"1.0000\"\n        data-duration=\"6.5000\"\n        data-track-index=\"2\"\n        style=\"left:0;top:0;width:1920px;height:1080px;visibility:hidden;opacity:0;\"\n      >\n        \u003C!-- paste the contents of public\u002Fcards\u002Fcard-01.html here -->\n      \u003C\u002Fdiv>\n\n      \u003C!-- Example: card-02 with zone=\"side-panel\" (split composition layout) → card on left half -->\n      \u003Cdiv\n        class=\"card-host clip\"\n        data-card-id=\"card-02\"\n        data-start=\"8.0000\"\n        data-duration=\"12.0000\"\n        data-track-index=\"2\"\n        style=\"left:0;top:0;width:960px;height:1080px;visibility:hidden;opacity:0;\"\n      >\n        \u003C!-- card-02 HTML -->\n      \u003C\u002Fdiv>\n\n      \u003C!-- ...one \"card-host clip\" per card with inline bounds matching resolveZoneBounds(card.zone)... -->\n\n      \u003Cscript src=\"vendor\u002Fgsap.min.js\">\u003C\u002Fscript>\n      \u003Cscript>\n        (function () {\n          \u002F\u002F count-up formatter helper\n          window.__fmt = function (v, fmt) {\n            if (typeof fmt === \"string\" && \u002F^\\.[0-9]+f$\u002F.test(fmt)) {\n              return Number(v).toFixed(Number(fmt.slice(1, -1)));\n            }\n            if (fmt === \",d\") return Math.round(v).toLocaleString();\n            return String(Math.round(v));\n          };\n\n          const tl = window.gsap.timeline({ paused: true });\n\n          \u002F\u002F ── Card lifecycle (one block per card) ──\n          \u002F\u002F Example for card-01 [1.0, 7.5] with kinetic-chars at +0.3, grow-x at +0.65:\n\n          \u002F\u002F Enter (fade in over 0.4s)\n          tl.set('.card-host[data-card-id=\"card-01\"]', { visibility: \"visible\" }, 1.0);\n          tl.fromTo(\n            '.card-host[data-card-id=\"card-01\"]',\n            { opacity: 0 },\n            { opacity: 1, duration: 0.4, ease: \"power2.out\" },\n            1.0,\n          );\n\n          \u002F\u002F Card-internal anims (compile each data-anim-* declaration here)\n          tl.from(\n            '.card[data-card-id=\"card-01\"] #card-01-title .char',\n            { opacity: 0, y: 8, scale: 0.8, duration: 0.5, ease: \"power2.out\", stagger: 0.04 },\n            1.3,\n          );\n          tl.fromTo(\n            '.card[data-card-id=\"card-01\"] #card-01-line',\n            { width: 0 },\n            { width: 420, duration: 0.5, ease: \"power2.out\" },\n            1.65,\n          );\n\n          \u002F\u002F Exit (fade out over 0.35s, ending at endSec)\n          tl.to(\n            '.card-host[data-card-id=\"card-01\"]',\n            { opacity: 0, duration: 0.35, ease: \"power2.in\" },\n            7.15,\n          );\n          tl.set('.card-host[data-card-id=\"card-01\"]', { visibility: \"hidden\" }, 7.5);\n\n          \u002F\u002F ── Video framing transitions ──\n          \u002F\u002F When the next card uses a different composition layout, animate the\n          \u002F\u002F video-wrapper to its new bounds. Example: card-01 = fullscreen\n          \u002F\u002F (video hidden behind), card-02 = split composition (zone=\"side-panel\"\n          \u002F\u002F → video on right, card on left).\n\n          \u002F\u002F Card-02 enters at 8.0s with the split composition. Animate video to\n          \u002F\u002F the right half during the card-01 → card-02 gap (between 7.5 and 8.0s).\n          tl.set(\"#video-wrap\", { className: \"video-wrapper framed\" }, 7.5);\n          tl.to(\n            \"#video-wrap\",\n            { left: 960, top: 0, width: 960, height: 1080, duration: 0.6, ease: \"power2.inOut\" },\n            7.5,\n          );\n\n          \u002F\u002F Card-02 enter — same pattern as card-01\n          tl.set('.card-host[data-card-id=\"card-02\"]', { visibility: \"visible\" }, 8.0);\n          tl.fromTo(\n            '.card-host[data-card-id=\"card-02\"]',\n            { opacity: 0 },\n            { opacity: 1, duration: 0.4, ease: \"power2.out\" },\n            8.0,\n          );\n          \u002F\u002F ...card-02 internal anims...\n\n          \u002F\u002F ── repeat for each card; if the NEXT card's layout differs,\n          \u002F\u002F    insert another tl.to('#video-wrap', ...) tween before its enter ──\n\n          window.__timelines = window.__timelines || {};\n          window.__timelines[\"talking-head-recut\"] = tl;\n        })();\n      \u003C\u002Fscript>\n    \u003C\u002Fdiv>\n  \u003C\u002Fbody>\n\u003C\u002Fhtml>\n```\n\n#### GSAP Statement Cheat Sheet\n\nCompile each `data-anim` attribute into a GSAP statement. Times are\n**absolute seconds** = card.startSec + data-anim-at, quantized to 1\u002Ffps.\nSelector is `.card[data-card-id=\"X\"] #elementId`.\n\n| data-anim                       | GSAP statement template                                                                                                                                                                                            |\n| ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |\n| `fade-in`                       | `tl.fromTo(SEL, { opacity: 0 }, { opacity: 1, duration: D, ease: 'power2.out' }, T);`                                                                                                                              |\n| `fade-out`                      | `tl.to(SEL, { opacity: 0, duration: D, ease: 'power2.in' }, T);`                                                                                                                                                   |\n| `slide-in` (from=left, dist=80) | `tl.fromTo(SEL, { opacity: 0, x: -80 }, { opacity: 1, x: 0, duration: D, ease: 'power2.out' }, T);`                                                                                                                |\n| `kinetic-chars` (pop)           | `tl.from(SEL + ' .char', { opacity: 0, y: 8, scale: 0.8, duration: D, ease: 'power2.out', stagger: S }, T);`                                                                                                       |\n| `count-up`                      | `(function(){const o={v:FROM};tl.to(o,{v:TO,duration:D,ease:'power2.out',onUpdate:function(){const el=document.querySelector(SEL);if(el)el.textContent=__fmt(o.v,'FMT');}},T);})();`                               |\n| `draw-path`                     | `(function(){const el=document.querySelector(SEL);if(el){const L=el.getTotalLength();tl.set(SEL,{strokeDasharray:L,strokeDashoffset:L},T);tl.to(SEL,{strokeDashoffset:0,duration:D,ease:'power2.inOut'},T);}})();` |\n| `grow-x` (target-w=W)           | `tl.fromTo(SEL, { width: 0 }, { width: W, duration: D, ease: 'power2.out' }, T);`                                                                                                                                  |\n| `grow-y` (target-h=H)           | `tl.fromTo(SEL, { height: 0 }, { height: H, duration: D, ease: 'power2.out' }, T);`                                                                                                                                |\n| `scale-pop`                     | `tl.fromTo(SEL, { opacity: 0, scale: 0.6 }, { opacity: 1, scale: 1, duration: D, ease: 'back.out(1.6)' }, T);`                                                                                                     |\n| `mask-reveal` (direction=left)  | `tl.fromTo(SEL, { clipPath: 'inset(0 100% 0 0)' }, { clipPath: 'inset(0 0 0 0)', duration: D, ease: 'power2.inOut' }, T);`                                                                                         |\n\nQuantize: `T = Math.round(absSec * fps) \u002F fps`. At 30fps the smallest\nstep is `1\u002F30 ≈ 0.0333s`; rounding to 4 decimals (`.toFixed(4)`) is fine\ninside the JS literal.\n\n#### Video Framing Reference (per `layout` value)\n\nThe selector for the video container is `#video-wrap`. Animate its\nbounds between cards using `tl.to('#video-wrap', { ...bounds }, T)`.\nInitial bounds should be set inline on the element to match card-01's\nlayout. Pick a transition duration of 0.5–0.7s with `ease: 'power2.inOut'`.\n\n**Decorative frames** (`clean` \u002F `hairline` \u002F `polaroid`) sit as a\n**sibling** of `#video-wrap` and follow it through layout transitions.\nSee\n[`references\u002Fframes\u002F`](references\u002Fframes\u002F) for each frame's placement\nHTML, suggested CSS, and which layouts it pairs with. Quick rule:\n`overlay` layout suppresses decorative frames (the full-bleed video\nclashes with chrome); PiP layouts already have their own pill treatment\n(border-radius + white ring + shadow), so add a decorative frame only on\ntop of `split` \u002F `stack`.\n\n**GSAP target lookup table** for `#video-wrap` per composition layout\n(landscape 1920×1080 — for portrait & 4:5 see `references\u002Flayouts\u002F*.html`\nwhich list all three ratios):\n\n| composition layout                   | typical card.zone | `#video-wrap` GSAP target                                                 | extra css class                            |\n| ------------------------------------ | ----------------- | ------------------------------------------------------------------------- | ------------------------------------------ |\n| `split`                              | `side-panel`      | `{ left: 960, top: 0, width: 960, height: 1080 }`                         | —                                          |\n| `stack`                              | `lower-third`     | `{ left: 14, top: 14, width: 1892, height: 548 }` (top 52%)               | —                                          |\n| `pip` (bottom-right)                 | `fullscreen`      | `{ left: 1480, top: 760, width: 400, height: 300 }`                       | `pip-pill` (border-radius + ring + shadow) |\n| `pip` (top-left)                     | `fullscreen`      | `{ left: 40, top: 40, width: 400, height: 300 }`                          | `pip-pill`                                 |\n| `overlay` (video full-bleed)         | `video-overlay`   | `{ left: 0, top: 0, width: 1920, height: 1080 }` (no change from default) | —                                          |\n| **hide video** (pure-graphic moment) | `fullscreen`      | `{ opacity: 0 }` (or move off-canvas)                                     | —                                          |\n\nTo toggle the pip-pill chrome (border-radius + white ring + drop shadow)\nwhen entering or leaving a pip moment:\n\n```js\n\u002F\u002F Enter pip — add chrome\ntl.set(\"#video-wrap\", { className: \"video-wrapper pip-pill\" }, T);\ntl.to(\n  \"#video-wrap\",\n  { left: 1480, top: 760, width: 400, height: 300, duration: 0.6, ease: \"power2.inOut\" },\n  T,\n);\n\n\u002F\u002F Leave pip — back to clean full-bleed\ntl.set(\"#video-wrap\", { className: \"video-wrapper\" }, T_NEXT);\ntl.to(\n  \"#video-wrap\",\n  { left: 0, top: 0, width: 1920, height: 1080, duration: 0.6, ease: \"power2.inOut\" },\n  T_NEXT,\n);\n```\n\n**Card-host bounds match the zone**. Resolve the card's `zone` into\npixel bounds using the table at the top of Step 6, then write those\ninto the card-host's inline `style=\"left:Xpx;top:Ypx;width:Wpx;\nheight:Hpx;...\"`. For `video-overlay` zone (overlay recipe), the\ncard-host fills the full canvas — your CSS inside `.card .root`\ndecides where the actual visible card sits.\n\n#### HyperFrames Layout \u002F Animation QA Rules\n\n- Build each card's static hero frame first: the moment where the card is fully visible and readable.\n- Confirm video, cards, subtitles\u002Fcaptions, and diagrams do not unintentionally overlap.\n- Confirm hidden video areas are clipped by the frame and not visible outside intended bounds.\n- Register one paused master timeline as `window.__timelines[\"talking-head-recut\"]`.\n- Build timelines synchronously at page load; no `async`, `setTimeout`, Promises, or media `play()` calls.\n- Do not use `Math.random()` or `Date.now()` in render paths.\n- Do not use `repeat: -1`; calculate finite repeats from the video duration.\n- Prefer GSAP transforms and opacity (`x`, `y`, `scale`, `rotation`, `opacity`) over layout properties (`top`, `left`, `width`, `height`) for motion.\n- Animate wrappers such as `#video-wrap`, not the video element dimensions directly.\n- Avoid animating the same property on the same element from multiple timelines at the same time.\n- Use `data-track-index`, not `data-layer`; use `data-duration`, not `data-end`.\n- Every timed element (`card-host`, sub-composition, etc.) MUST include `class=\"clip\"` alongside its own classes — e.g. `class=\"card-host clip\"`. The HyperFrames runtime uses `.clip` to gate visibility to the `data-start … data-start+data-duration` window. Without it the element is visible for the whole video (lint: `timed_element_missing_clip_class`).\n- For body \u002F global `font-family`, list **concrete font names** (`'Inter', 'Caveat', …`) — not a CSS variable like `var(--font-family)`. The HyperFrames font resolver doesn't expand CSS vars during static analysis (lint: `font_family_without_font_face`). Cards may still use `var(--font-family)` internally since their `@font-face` declarations are loaded.\n\n### 10. Render to MP4\n\n```bash\ncd \"$WORK_DIR\"\nPRODUCER_BROWSER_GPU_MODE=hardware npx hyperframes render public \\\n  --skill=talking-head-recut \\\n  -o output.mp4 \\\n  --fps 30\n```\n\n`hyperframes render \u003Cdir>` reads `\u003Cdir>\u002Findex.html` and produces the MP4.\nThe canonical composition keeps the visual `\u003Cvideo>` muted and mounts the same\nsource as the root `#source-audio` track, so the rendered MP4 preserves the\ntalking-head audio without a manual remux. This uses a separate audio track\nrather than `data-has-audio=\"true\"` so its volume and ducking remain independently\ncontrollable on the timeline.\nThe flag `PRODUCER_BROWSER_GPU_MODE=hardware` (or `--browser-gpu`) is\nstrongly recommended on macOS — software-only Chrome rendering times out\non most laptops.\n\nFor a sanity check before the full render, capture a single frame at a\nspecific timestamp:\n\n```bash\nnpx hyperframes snapshot public --at 5    # → public\u002Fsnapshots\u002Fframe-00-at-5s.png (a single --at ignores --out)\n```\n\n### 11. Report Results\n\nTell the user:\n\n- Work directory path\n- `storyboard.json` (the card outline you designed)\n- `public\u002Fcards\u002F*.html` (one HTML per card)\n- `public\u002Findex.html` (the assembled composition)\n- `output.mp4` (the final video)\n- ASR provider used\n- Card count + how you chose them (in 1 sentence)\n- Any missing keys or quality caveats\n\n**Optional live preview (on request only).** The clip plays unchanged inside `public\u002Findex.html` with the overlays on top, so it previews faithfully. **Don't open it during the run.** When the user asks, start a long-lived server **after** render and report the URL:\n\n```bash\n(cd \"$WORK_DIR\u002Fpublic\" && npx hyperframes preview)   # or `npx hyperframes play` for a shareable link\n```\n\nDo not delete the work directory unless the user asks.\n",{"description":7},"main","58843b159d7a4050d6b785a98b908cce505c0b92","agent\u002Fskills\u002Ftalking-head-recut",0,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent","https:\u002F\u002Fgithub.com\u002FReihantt6","reihsns.web.id","2026-08-19 20:08:18",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add Reihantt6\u002Fsns-myagent --skill talking-head-recut",{"tool":38,"command":39},"bunx","bunx skills add Reihantt6\u002Fsns-myagent --skill talking-head-recut",{"tool":41,"command":42},"pnpm","pnpm dlx skills add Reihantt6\u002Fsns-myagent --skill talking-head-recut","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002FSKILL.md",true,[46,52,55,59,63,67,71,77,83,87,91,95,99,103,107,111,115,119,123,127,131,135,139,143,147,151],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"assets\u002Ffonts\u002FCaveat-400-latin.woff2","ASSET","资产文件",74932,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Fassets\u002Ffonts\u002FCaveat-400-latin.woff2",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":54},"assets\u002Ffonts\u002FCaveat-700-latin.woff2","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Fassets\u002Ffonts\u002FCaveat-700-latin.woff2",{"path":56,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":57,"sourceHtmlUrl":58},"assets\u002Ffonts\u002FInter-400-latin.woff2",23692,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Fassets\u002Ffonts\u002FInter-400-latin.woff2",{"path":60,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":61,"sourceHtmlUrl":62},"assets\u002Ffonts\u002FInter-700-latin.woff2",24352,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Fassets\u002Ffonts\u002FInter-700-latin.woff2",{"path":64,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":65,"sourceHtmlUrl":66},"assets\u002Ffonts\u002FLXGWWenKaiTC-400-latin.woff2",19380,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Fassets\u002Ffonts\u002FLXGWWenKaiTC-400-latin.woff2",{"path":68,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":69,"sourceHtmlUrl":70},"assets\u002Ffonts\u002FVirgil.woff2",61248,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Fassets\u002Ffonts\u002FVirgil.woff2",{"path":72,"resourceType":73,"resourceTypeDesc":74,"sizeBytes":75,"sourceHtmlUrl":76},"media-contract.test.mjs","SCRIPT","脚本",1350,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Fmedia-contract.test.mjs",{"path":78,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":81,"sourceHtmlUrl":82},"NOTICE.md","REFERENCE","参考文档",1826,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002FNOTICE.md",{"path":84,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":85,"sourceHtmlUrl":86},"references\u002FDESIGN_INDEX.md",10171,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002FDESIGN_INDEX.md",{"path":88,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":89,"sourceHtmlUrl":90},"references\u002Fframes\u002Fclean.html",2626,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fframes\u002Fclean.html",{"path":92,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":93,"sourceHtmlUrl":94},"references\u002Fframes\u002Fhairline.html",5884,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fframes\u002Fhairline.html",{"path":96,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":97,"sourceHtmlUrl":98},"references\u002Fframes\u002Fpolaroid.html",5349,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fframes\u002Fpolaroid.html",{"path":100,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":101,"sourceHtmlUrl":102},"references\u002Flayouts\u002Foverlay.html",5463,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Flayouts\u002Foverlay.html",{"path":104,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":105,"sourceHtmlUrl":106},"references\u002Flayouts\u002Fpip.html",4650,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Flayouts\u002Fpip.html",{"path":108,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":109,"sourceHtmlUrl":110},"references\u002Flayouts\u002Fsplit.html",3805,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Flayouts\u002Fsplit.html",{"path":112,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":113,"sourceHtmlUrl":114},"references\u002Flayouts\u002Fstack.html",3458,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Flayouts\u002Fstack.html",{"path":116,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":117,"sourceHtmlUrl":118},"references\u002Fstyles\u002Facademic.html",6043,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Facademic.html",{"path":120,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":121,"sourceHtmlUrl":122},"references\u002Fstyles\u002Faudit.html",6182,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Faudit.html",{"path":124,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":125,"sourceHtmlUrl":126},"references\u002Fstyles\u002Feditorial.html",6586,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Feditorial.html",{"path":128,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":129,"sourceHtmlUrl":130},"references\u002Fstyles\u002Fgeom.html",5776,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Fgeom.html",{"path":132,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":133,"sourceHtmlUrl":134},"references\u002Fstyles\u002Fminimal.html",4142,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Fminimal.html",{"path":136,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":137,"sourceHtmlUrl":138},"references\u002Fstyles\u002Fspotlight.html",4941,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Fspotlight.html",{"path":140,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":141,"sourceHtmlUrl":142},"references\u002Fstyles\u002Fswiss.html",6053,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Fswiss.html",{"path":144,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":145,"sourceHtmlUrl":146},"references\u002Fstyles\u002Fterminal.html",6957,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Fterminal.html",{"path":148,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":149,"sourceHtmlUrl":150},"references\u002Fstyles\u002Fwhiteboard.html",5829,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Fwhiteboard.html",{"path":152,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":153,"sourceHtmlUrl":154},"references\u002Fstyles\u002Fxhs.html",7210,"https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Ftalking-head-recut\u002Freferences\u002Fstyles\u002Fxhs.html",[156,163,170,177,184,191],{"skillNo":157,"name":158,"nameZh":6,"description":159,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":160,"categories":161,"keywords":162,"publishedAt":19},"01M77XDCYZ407XYMF2T4P6F7VT","youtube-transcript","Extract transcripts from YouTube videos. Use when the user asks for a transcript, subtitles, or captions of a YouTube video and provides a YouTube URL (youtube.com\u002Fwatch?v=, youtu.be\u002F, or similar). Supports output with or without timestamps.","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fyoutube-transcript\u002FSKILL.md",[],[],{"skillNo":164,"name":165,"nameZh":6,"description":166,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":167,"categories":168,"keywords":169,"publishedAt":19},"01M77XD7PBC98J6MXGJJ2TJ8DX","youtube-thumbnail-design","YouTube thumbnail design with specific dimensions, contrast rules, and mobile preview optimization. Covers safe zones, text placement, face expression psychology, and A\u002FB testing. Use for: YouTube thumbnails, video cover images, click-through optimization. Triggers: youtube thumbnail, thumbnail design, video thumbnail, click through rate, ctr optimization, youtube cover, video cover image, thumbnail maker, thumbnail tips, youtube design, video preview image","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fyoutube-thumbnail-design\u002FSKILL.md",[],[],{"skillNo":171,"name":172,"nameZh":6,"description":173,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":174,"categories":175,"keywords":176,"publishedAt":19},"01M77XD2541C8HXJ6RHYZP4NGJ","youtube-short-maker","Make YouTube Shorts that hold attention with Pexo — vertical 9:16, fast-paced, hook in the first 3 seconds. Describe your topic and Pexo writes the script, generates the shots, picks the models, and assembles a finished Short with music and captions. Use for YouTube Shorts: \"youtube short\", \"make a youtube short\", \"shorts video\", \"short-form video\". NOT for long-form or landscape video.","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fyoutube-short-maker\u002FSKILL.md",[],[],{"skillNo":178,"name":179,"nameZh":6,"description":180,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":181,"categories":182,"keywords":183,"publishedAt":19},"01M77XCXKVSYK7KTZ3PQBVVGFD","xcode-project-setup","Safely modifies Xcode projects (.pbxproj) to add Swift Packages and link files. Use this skill whenever an iOS project needs dependencies installed (e.g. Firebase, Alamofire).","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fxcode-project-setup\u002FSKILL.md",[],[],{"skillNo":185,"name":186,"nameZh":6,"description":187,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":188,"categories":189,"keywords":190,"publishedAt":19},"01M77XCR1DV8ZQK2NHWKVY4EWX","x402","Set up Browser Use Cloud payments with x402 — pay per request from a crypto wallet (USDC on Base mainnet), no signup or API key. Two setups it works out up front — \"just use it\" (set up a wallet so you or Claude Code can run cloud browser tasks paid from the wallet — Claude writes and runs throwaway scripts, nothing touches your codebase) or \"build it in\" (install the SDK and write the key + code into your project). Walks through wallet setup, funding, .env, and a ~$1 test run. Use when the user asks about x402, pay-per-use, USDC payments, or wants Browser Use Cloud without an API key. For the free-tier signup (reverse-CAPTCHA → API key), use `browser-use cloud signup` or the `cloud` skill instead.","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fx402\u002FSKILL.md",[],[],{"skillNo":192,"name":193,"nameZh":6,"description":194,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":195,"categories":196,"keywords":197,"publishedAt":19},"01M77XCMD3R3ZBZQQSW6E0793B","x-twitter-scraper","Build GitHub Copilot workflows with Xquik X API SDKs, REST endpoints, hosted Apify Actor runs, MCP tools, TweetClaw OpenClaw plugin installs, signed webhooks, tweet search, user lookup, follower exports, media actions, and agent automation.","https:\u002F\u002Fgithub.com\u002FReihantt6\u002Fsns-myagent\u002Fblob\u002Fmain\u002Fagent\u002Fskills\u002Fx-twitter-scraper\u002FSKILL.md",[],[]]