Update to Astria

{
  "name": "storyboard",
  "description": "Turn a rough scene, image prompt, or reference-led visual idea into a text-only cinematic video storyboard. Use for /storyboard, \"turn this into a cinematic scene\", \"storyboard this video\", or direct text-to-video planning. Write the finished shot sequence into video_prompt and clear the image prompt; never generate a storyboard-grid image.",
  "included_files": [],
  "skill_md_contents": "---\nname: storyboard\ndescription: Turn a rough scene, image prompt, or reference-led visual idea into a text-only cinematic video storyboard. Use for /storyboard, \"turn this into a cinematic scene\", \"storyboard this video\", or direct text-to-video planning. Write the finished shot sequence into video_prompt and clear the image prompt; never generate a storyboard-grid image.\n---\n\n# Video Storyboard\n\nTurn the current draft into a cinematic shot sequence written directly in the\nvideo prompt. Never generate a 4x4 artboard, storyboard image, contact sheet, or\nfirst-frame image.\n\nAlways write the finished storyboard and all generation prompt text in English,\neven when the user communicates or supplies the draft in another language. The\nsurrounding conversation may remain in the user's language. Translate\nnon-English draft content into English while preserving its meaning and all\nreference tokens exactly.\n\n## Use the current draft\n\nRead the current image prompt, reference tokens, aspect ratio, and any existing\nvideo prompt from page context. Also read `image_reference_urls` as an ordered\nsequence of raw visual waypoints. Treat a handoff such as \"Turn this into a\ncinematic scene\" as permission to make the missing creative choices when the\ndraft already contains a reference, scene, and general idea of the frame. Do\nnot ask the user to repeat those details.\n\nDiscard every reference whose tune `name` is `pose`, including its complete\n`<lora:...> pose` or `<faceid:...> pose` mention. Never copy a `pose` tune into\nthe storyboard or use its reference image for the video. This applies even when\ncopying an otherwise completed prompt verbatim. Do not remove ordinary prose\nthat describes a subject's pose, stance, or movement.\n\nPreserve every other `<lora:...>` and `<faceid:...>` token exactly. Keep the\nsame character, products, wardrobe, location, time of day, lighting, and grade\nthroughout the sequence unless the requested story deliberately changes one.\nPreserve every raw image reference exactly and in the same order. Do not turn\nraw references into tunes or embed their URLs in the storyboard text. A raw\nreference may be a public HTTP(S) URL or an absolute `/workspace/...` path from\nan attached file; keep either form unchanged and never rewrite a workspace path\nas `file://` or invent a public URL for it.\n\n## Write the storyboard\n\nDefault to 16 numbered shots for a 15-second video unless the user explicitly\nrequests another supported duration:\n\n- Give each shot one camera scale, one subject, and one filmable action.\n- Do not repeat a camera scale twice in a row. Rotate among extreme close-up,\n  close-up, medium, long, and extreme long shots.\n- Vary angles with front, back, profile, low-angle, top-down, and\n  over-the-shoulder views where useful.\n- Use shots 1-3 to establish the hero, environment, and motion; shots 4-13 for\n  the action montage; and shots 14-16 for a product/detail beat and a closing\n  wide shot.\n- Put the global light, color grade, grain, and atmosphere in shot 1, then keep\n  them continuous.\n- Use the identical reference token whenever its subject appears. Chain\n  continuity with phrases such as \"the same woman\", \"she\", and \"her\".\n- When ordered raw image references are present, make the action progress from\n  reference 1 through the final reference in order, describing filmable visual\n  transitions between them rather than treating them as unrelated examples.\n\nWrite only the numbered shots in the finished video prompt. Do not add a grid\nheader or describe storyboard tiles.\n\n## Apply it to the composer\n\nChoose the video model before calling `present_generation`:\n\n- Preserve an explicit video model already selected in the current draft.\n- Otherwise, reuse the Seedance 2.5 or Seedance 2 model and resolution established\n  by the user's prior videos.\n- If prior videos do not establish either default, ask the user to choose between\n  Seedance 2.5 and Seedance 2. Do not silently use the composer's global default.\n\nCall `present_generation` exactly once with the complete current generation\ndraft. Put the completed sequence in `video_prompt`, set `text` to the empty\nstring, discard all `pose` tune references, and preserve the remaining fields\nfrom Current generation draft JSON, including the ordered\n`image_reference_urls` array. Set `video_duration` to `15` unless the user\nexplicitly requested another supported duration. Set `video_model` to the\nexplicit, established, or user-selected model from the rules above. This writes\nthe sequence directly to video mode with no image/first-frame prompt. Do not\nrepeat the storyboard in assistant text and do not emit an `ASTRIA_PROMPT` or\n`ASTRIA_VIDEO_PROMPT` command.\n\nIf the current image prompt already contains the completed storyboard or the\nuser asks to move the current prompt into video mode, remove any `pose` tune\nreference, then put everything else into the `present_generation`\n`video_prompt` field and set `text` to the empty string. Preserve English\ncontent verbatim. If the content is not in English, translate it faithfully\ninto English without summarizing, prefixing, trimming, or otherwise rewriting\nit beyond translation.\n\n## Generate video\n\nWhen the user explicitly asks to generate, pass the exact approved, pose-tune-\nfree storyboard as `--video-prompt`, omit `--text` entirely, and default to a\n15-second duration unless the user explicitly requested another supported\nduration:\n\n- If `prompt.text` still contains the content and `video_prompt` is empty,\n  discard any `pose` tune reference, move the remaining `prompt.text` into\n  `video_prompt`, and clear `prompt.text` before generation. Preserve English\n  content byte-for-byte. Translate non-English content faithfully into English;\n  do not expand, summarize, prefix, trim, or otherwise rewrite it beyond\n  translation during this generation step.\n- If `video_prompt` is already populated, discard any `pose` tune reference and\n  use English content as-is; translate non-English content faithfully into\n  English before generation.\n\n```bash\nastria video --video-model \"<the chosen Seedance 2.5 or Seedance 2 model>\" \\\n  --duration 15 --num-images 1 \\\n  --image-reference \"<reference 1>\" --image-reference \"<reference 2>\" \\\n  --video-prompt \"<the exact approved storyboard>\"\n```\n\nInclude one `--image-reference` for every raw reference, in its original order.\nOmit those options when the draft has no raw image references. Use either all\nlocal paths or all URLs in a single command.\n\nReplace `15` only when the user explicitly requested another supported\nduration.\n\nUse the current aspect ratio when it is available. Raw `--image-reference`\nwaypoints are not first-frame prompts. Do not create or pass an artboard image,\nan input image, or an image/first-frame prompt unless the user explicitly asks\nfor one.\n"
}