Update to Crealix

{
  "description": "Make a narrated video with Crealix over real stock footage, a clip for every sentence, from a topic or the user's script. Use when the user wants real footage instead of AI images, for travel, finance, business, health, news-style explainers or motivation. Also swaps clips afterwards.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 412
    }
  ],
  "name": "stock-footage-video",
  "skill_md_contents": "---\nname: stock-footage-video\ndescription: Make a narrated video with Crealix over real stock footage, a clip for every sentence, from a topic or the user's script. Use when the user wants real footage instead of AI images, for travel, finance, business, health, news-style explainers or motivation. Also swaps clips afterwards.\n---\n\n# Stock footage video\n\nA narrator reads the script while real stock clips play under it, each one\nmatched to its sentence, with captions and music. Every sentence becomes a\nfootage search, so the writing decides what the viewer sees.\n\n## 1. Read the guide\n\nCall `get_creation_guide('stock', durationSec, language)`: the words budget for\nthis length, the cast (voices, captions, music) and the price. Stock footage is\nincluded; the user pays for the narration and, later, the render.\n\n## 2. Write for footage\n\n- About 2.5 words per second (the guide gives the figure). Use hybrid mode and\n  write it yourself. Use crealix mode only when the user asks Crealix to write it.\n- Hook in the first sentence. No greetings.\n- One idea per sentence, and each sentence describes something a camera could\n  film: people doing things, places, objects, weather, cities, nature.\n  \"A trader stares at red numbers on six screens\" finds footage; \"the market\n  is uncertain\" does not.\n- Keep abstract lines rare and next to concrete ones, so the footage still makes\n  sense.\n- Plain text, no headings or stage directions. The words are read as written.\n- Facts, figures and claims: only what you know is true.\n\n## 3. Cast it\n\nChoose from the cast what suits this video: its topic, tone, audience and\nplatform.\n- Voice and delivery (`voiceSettings` adapted from `voiceDelivery`).\n- Captions and music under the voice.\n- `cover` for short-form, and `voiceEffect` only when it serves the story.\n- 9:16 for TikTok, Reels and Shorts; 16:9 for YouTube; 1:1 when asked.\n\n## 4. Propose and create\n\n`check_plan` until it is valid and fits the length, then one `validate_plan`, with\ntwo or three lines on why your choices suit this video. On a yes in the chat,\n`create_video` with the `proposalId`. A press on the card's Create button has\nalready created it.\n\n## 5. Improve the footage\n\nWhen it is ready (`get_status`), the card plays the draft. If a scene shows the\nwrong thing:\n1. `search_stock_media` with a concrete query for that line, at the video's shape.\n2. `import_media_from_url` with the chosen clip's `videoUrl`.\n3. `edit_video({ scenes })` to put the imported clip on that scene number.\n\n`edit_video` also changes the title, captions, music and mix. The narrator stays\nthe one chosen at creation. Edits are free and `undo_last_change` reverses them.\nRender only when they ask: the card's Render button, or `render_project`.\n"
}