sources_v3.json 19.2 KB
{
  "task": "T-mtb58zsc",
  "agent": "innovator-hunt (Hunter)",
  "round": "retrain v3 HUNT",
  "crawl_date": "2026-08-27",
  "fetched_at_utc": "2026-08-27T06:51:49Z",
  "method": "curl 直读 raw.githubusercontent.com 全文,字节数与 api.github.com git/trees/main 返回逐一核对一致;行号为文件内 1-based 行号",
  "notes": [
    ".agents/skills/h3-prompt-writing/* 与 .claude/skills/h3-prompt-writing/* 为同字节副本(2177/15773/23553),本清单引用 .claude 路径",
    "platform.minimax.io / platform.minimaxi.com 本轮不可达(WebFetch 域校验被拒、curl 三次超时),开放平台 API schema 未核实"
  ],
  "sources": [
    {
      "id": "s01",
      "title": "MiniMax-H3 官方技能总说明 SKILL.md",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/.claude/skills/h3-prompt-writing/SKILL.md",
      "bytes": 2177,
      "sha256_16": "a48f9db1231cebe6",
      "crawl_time": "2026-08-27T06:44:00Z",
      "quotes": [
        { "line": 14, "text": "4. Preserve the exact field names, section order, labels, and timing notation from the selected guide." },
        { "line": 27, "text": "Ref2VA rewrites use `subject_definitions`, `summary`, `retention_analysis`, `detailed_description`, `overall_soundscape`, and `non_diegetic_music` in that order. Reference labels stay consistent across all sections." },
        { "line": 33, "text": "- Write rewrite sections in English; preserve dialogue, lyrics, and visible scene text in their original language." }
      ],
      "used_for": ["需求(2) 读法A风险依据", "补充B 台词原语言保真"]
    },
    {
      "id": "s02",
      "title": "官方基础模式规范 base-en.txt(全文 223 行到手)",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/.claude/skills/h3-prompt-writing/references/base-en.txt",
      "bytes": 15773,
      "sha256_16": "2cfebc096a6e0837",
      "crawl_time": "2026-08-27T06:45:30Z",
      "quotes": [
        { "line": 54, "text": "`<Picture 1>` is the actual first frame of the video at 0.00 seconds and belongs to `[Shot 1]`. The description should first establish the style, subjects, composition, and scene anchors in the image, then describe the next action. Character identity, clothing, colors, key objects, and spatial relationships should remain consistent." },
        { "line": 92, "text": "For ordinary cuts, use `the camera cuts to`, `the shot cuts to`, `the shot transitions to`, `the shot changes to`, or `the shot switches to`. When explicitly requested by the user, cross-dissolve, fade, or wipe may also be used. A cut should introduce new information about the subject, space, state, viewpoint, or time." },
        { "line": 127, "text": "Subjects who speak, sing, or produce an off-screen human voice use stable IDs such as `(S1)` and `(S2)`. When multiple already-numbered speakers speak or sing together, use a compound ID such as `(S1,S2)`. A speaker keeps the same ID across shots; characters who never vocalize receive no speaker ID." },
        { "line": 129, "text": "When a speaker first appears, provide enough information from the visual and audio context to establish a stable identity, such as character type, age, gender, whether the person is on-screen, pitch, timbre, speaking rate, or accent. Place the speaker's identifying phrase, ID, action, and delivery outside `<d>`. Inside `<d>`, include only the language tag and the actual user-provided spoken content. Preserve every original word and punctuation mark verbatim; do not translate or rewrite them." },
        { "line": 136, "text": "For voiceover, use the exact phrase `says in an off-screen voiceover`. Immediately after every voiceover `<d>` block, state that the corresponding on-screen character's lips remain closed:" },
        { "line": 139, "text": "The man (S1) says in an off-screen voiceover: <d>[English] I still remember that road.</d> while his lips remain completely closed." },
        { "line": 172, "text": "With no reference image, construct the complete timeline directly from the text. You may add scene, character, action, and sound details that remain consistent with the user's intent." }
      ],
      "used_for": ["盲区① 双说话人编号", "盲区④ 画外音", "盲区⑥ 最接近的道具一致性条款", "补充A 原故事情节最接近条款", "§5.6 标点作用域·verbatim 一侧"]
    },
    {
      "id": "s03",
      "title": "官方 Ref2VA 规范 ref-en.txt(全文 342 行到手)",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/.claude/skills/h3-prompt-writing/references/ref-en.txt",
      "bytes": 23553,
      "sha256_16": "1e574f356716ad55",
      "crawl_time": "2026-08-27T06:46:10Z",
      "quotes": [
        { "line": 5, "text": "Write all six rewrite sections in English. Preserve the original language only for dialogue and lyrics inside `<d>` and for text visibly present in the scene." },
        { "line": 26, "text": "Full-reference rewrites use four types of labels to identify the source and role of referenced content:" },
        { "line": 28, "text": "| `<Subject N>` | Visible content abstracted from reference assets that can be reused or modified in the target video |" },
        { "line": 29, "text": "| `<Picture N>` | A reference image used as a concrete target frame or shot-planning anchor |" },
        { "line": 30, "text": "| `<Video N>` | A reference video that provides an editing source, continuation starting point, or whole-video temporal structure |" },
        { "line": 31, "text": "| `<Audio N>` | An audio signal that is copied or referenced |" },
        { "line": 37, "text": "subject_definitions` defines each piece of referenced content that must be tracked separately later, such as a person, an environment, a source video's structure, or an audio track. Give each item its own line and explain what its label denotes, its reference role, and the main features to follow; name the corresponding source asset when its provenance needs to be made explicit. If `<Picture N>` or `<Video N>` only identifies the source of another referenced item and will not be analyzed or used separately later, cite it inside that item's definition without adding a separate line." },
        { "line": 41, "text": "`<Subject N>` is used for reusable visible content, including:" },
        { "line": 43, "text": "- People, animals, or objects" },
        { "line": 44, "text": "- Scenes, backgrounds, or environments" },
        { "line": 48, "text": "It represents a content unit that will actually be used in the target video, rather than the source file itself. One subject may be defined by multiple reference assets, and one reference asset may provide multiple subjects." },
        { "line": 62, "text": "Use a standalone `<Picture N>` when the reference image itself serves as a shot's first frame, keyframe, last frame, edited keyframe, or composition anchor:" },
        { "line": 68, "text": "If an image is used only to define a character, scene, costume, or style, do not create a standalone picture entry. Instead, cite the image source inside the corresponding `<Subject N>` definition." },
        { "line": 73, "text": "<Picture 3> is a storyboard reference for [Shot 1] and [Shot 2], defining their viewpoint, subject placement, and shot order." },
        { "line": 92, "text": "`<Audio N>` represents a standalone audio asset or an enabled synchronized audio track from a reference video. Common uses include:" },
        { "line": 100, "text": "When an `<Audio N>` explicitly corresponds to a target speaker, reuse that speaker's global ID in the definition: write `<Subject N> (Sx)` when the speaker maps to a defined subject, or use a stable voice description followed by `(Sx)` otherwise. The ID comes from the target video's global speaker order and is not independently assigned or renumbered in the audio definition." },
        { "line": 110, "text": "`<Video N>` and `<Audio N>` are numbered independently. Each index indicates only the label's order within its own category and does not encode a pairing between the two categories." },
        { "line": 112, "text": "An ordinary reference video does not create `<Audio N>` merely because the file contains sound." },
        { "line": 165, "text": "`<Subject N>`, `<Picture N>`, and `<Video N>` use the following relationship markers. These markers are fixed English values in the output format:" },
        { "line": 168, "text": "| `weak_reference` | Only broad similarity in style, category, composition, or atmosphere is retained |" },
        { "line": 207, "text": "Choose each relationship marker only within the reference role already defined for that label in `subject_definitions`. Do not treat newly added actions, backgrounds, or plot events in the target video as losses of reference fidelity." },
        { "line": 246, "text": "At the first clear appearance of an important `<Subject N>`, describe its referenced characteristics, position in the frame, and current action within what is actually visible in the shot." },
        { "line": 266, "text": "If the same subject speaks off-screen, keep the same form and mark it as `off-screen`." },
        { "line": 274, "text": "When dialogue, narration, or lyrics from reference audio are directly reused, or when the input prompt explicitly requests their reperformance, preserve the exact source words and original language inside `<d>`. Write `[unclear]` for unintelligible spans instead of guessing or paraphrasing them. Standardize punctuation to the basic written marks needed to express the sentence, such as `,`, `.`, `?`, and `!`; remove repeated tildes, emoji, bullets, and repeated or decorative punctuation. End complete statements, questions, and exclamations with `.`, `?`, or `!` respectively before `</d>`." },
        { "line": 278, "text": "Assign `(Sx)` once according to the order of actual vocal events in the target video. Reuse the corresponding ID at every actual vocal event in `detailed_description`; an `<Audio N>` definition bound to a target speaker in `subject_definitions` also reuses the same `(Sx)` but never assigns a new one independently. Do not write `(Sx)` in `retention_analysis`." },
        { "line": 312, "text": "<Subject 1> is the coffee-shop environment in <Picture 1>, featuring an exposed brick wall, an orange tufted sofa with patterned pillows, a neon sign, and a wooden coffee table." },
        { "line": 313, "text": "<Subject 2> is the fluffy white Samoyed in <Picture 2>, <Picture 3>, and <Picture 4>, with thick white fur, pointed ears, a dark nose, and a curved tail." },
        { "line": 320, "text": "<Subject 1> (appears in [Shot 1], [Shot 2], [Shot 3]): fully_preserved - the exposed brick wall, orange tufted sofa, patterned pillows, neon sign, and wooden coffee table are retained." }
      ],
      "used_for": ["需求(1) 无 object/background 标签的反证", "需求(2) 读法B 映射依据", "需求(3) 立绘规则通用性", "盲区② Audio N 锚定", "新发现 §5.6 标点作用域分叉"]
    },
    {
      "id": "s04",
      "title": "官方仓库 README.md(含模型表、conditions[].uri 说明、开放平台 API 入口)",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/README.md",
      "bytes": 40135,
      "sha256_16": "7964f3044d3c7efa",
      "crawl_time": "2026-08-27T06:43:20Z",
      "quotes": [
        { "line": 58, "text": "- Global: [platform.minimax.io](https://platform.minimax.io/docs/api-reference/video-generation-v2-create) | CN: [platform.minimaxi.com](https://platform.minimaxi.com/docs/api-reference/video-generation-v2-create)" },
        { "line": 85, "text": "| H3-Base-Ref2VA | Omni-reference mode | Supports multi-modal reference inputs: <br><br>- **Images:** ≤ 9 images <br>- **Videos:** ≤ 3 clips; each clip must be 2–15 seconds long; total duration ≤ 15 seconds <br>- **Audio:** ≤ 3 clips; each clip must be 2–15 seconds long; total duration ≤ 15 seconds <br>- **Mixed inputs:** Maximum number of files across all input types is 12 |" },
        { "line": 270, "text": "#### Using a local image/video instead of a remote URL" },
        { "line": 272, "text": "The reproducible scripts above reference `conditions[].uri` values hosted on a public CDN so they work out of the box, but `uri` is not limited to `http(s)://`. For local testing you can point it at a `file://` path instead — for example, to swap the FL2VA keyframe for your own image:" },
        { "line": 285, "text": "The path is resolved by the SGLang server process, not by the machine running `curl`, so it must point to a file the server can actually see:" }
      ],
      "used_for": ["需求(1) 线上 URL 传入机制", "素材上限补齐(合计≤12)"]
    },
    {
      "id": "s05",
      "title": "官方 Ref2VA 可复现请求脚本(video+audio 条件、role=reference)",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/scripts/readme/reproducible-768p-ref2va-request.sh",
      "bytes": 5047,
      "sha256_16": "73ea0e5cc94440b2",
      "crawl_time": "2026-08-27T06:47:05Z",
      "quotes": [
        { "line": 11, "text": "{\n  \"task\": \"ref2va\",\n  \"prompt\": \"subject_definitions:\\n<Subject 1> is the young man with short wavy blonde hair, ...\",\n  \"conditions\": [\n    {\n      \"type\": \"video\",\n      \"uri\": \"https://cdn.hailuoai.com/prod/hailuo_demo/testsets/h3_promo_eval_ref2va/gallery/sr_v2p26_trio_seed42_20260724/inputs/297573323635_00_%E8%A7%86%E9%A2%911_YnyRbxEwio_video_20260525_163755_1927e9d3.mp4\",\n      \"role\": \"reference\"\n    },\n    {\n      \"type\": \"audio\",\n      \"uri\": \"https://cdn.hailuoai.com/prod/hailuo_demo/testsets/h3_promo_eval_ref2va/gallery/sr_v2p26_trio_seed42_20260724/inputs/f463d523c5ce_01_%E9%9F%B3%E9%A2%911_RSLcbpzJPo_6%E6%9C%885%E6%97%A5(1).mp3\",\n      \"role\": \"reference\"\n    }\n  ],\n  \"target\": {\n    \"short_edge\": 768,\n    \"aspect_ratio\": \"auto\",\n    \"duration_seconds\": 5\n  },\n  \"seed\": 0\n}" }
      ],
      "used_for": ["需求(1):线上 https URL 直接放 conditions[].uri;官方无 references 数组"]
    },
    {
      "id": "s06",
      "title": "官方 FL2VA 可复现请求脚本(image 条件、role=keyframe + frame_index)",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/scripts/readme/reproducible-768p-fl2va-request.sh",
      "bytes": 5718,
      "sha256_16": "82d1c06897da37e5",
      "crawl_time": "2026-08-27T06:47:40Z",
      "quotes": [
        { "line": null, "note": "conditions 块(图片用例)", "text": "\"conditions\": [\n  {\n    \"type\": \"image\",\n    \"uri\": \"https://cdn.hailuoai.com/prod/hailuo_demo/testsets/H3_AA_I2VA/gallery/sr_v17_variants_seed42_43_20260724/inputs/4a3a90bf9100_KDmcbkhzYo5sjjxr9FqcVmWVnzb.png\",\n    \"role\": \"keyframe\",\n    \"frame_index\": 0\n  }\n]" }
      ],
      "used_for": ["需求(1) 图片条件的 role 取值证据;同时暴露「图片+role=reference」官方案例缺失(存疑 §7.2)"]
    },
    {
      "id": "s07",
      "title": "Ref2VA processor chat_template.json(图像自动编号 Picture N 的机制性证据)",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/Ref2VA/processor/chat_template.json",
      "bytes": 5499,
      "sha256_16": "5c72a170d2a4a1a3",
      "crawl_time": "2026-08-27T06:48:10Z",
      "quotes": [
        { "line": null, "text": "{%- if content.type == 'image' or 'image' in content or 'image_url' in content %}\n  {%- set image_count.value = image_count.value + 1 %}\n  {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}\n  <|vision_start|><|image_pad|><|vision_end|>\n{%- elif content.type == 'video' or 'video' in content %}\n  {%- set video_count.value = video_count.value + 1 %}\n  {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}" }
      ],
      "used_for": ["需求(1)/(3):Picture 编号 = 素材顺序,与是否构图锚点无关(knowhow_v3 §2.4)"],
      "evidence_class": "机制性证据(实现层面推演依据),不是写作规范原文"
    },
    {
      "id": "s08",
      "title": "仓库全量路径清单 GitHub API(清点提示词相关文件、核对字节数)",
      "url": "https://api.github.com/repos/MiniMax-AI/MiniMax-H3/git/trees/main?recursive=1",
      "bytes": null,
      "crawl_time": "2026-08-27T06:36:00Z",
      "quotes": [
        { "line": null, "note": "摘要", "text": "tree 含 .agents/skills 与 .claude/skills 两份同字节 h3-prompt-writing 副本(SKILL.md 2177 / base-en.txt 15773 / ref-en.txt 23553);skills/ 下另有 8 个风格类视频生成 skill(minimalist-product-ad-generator 等);其余为模型权重目录 FL2VA//Ref2VA/ 等。无其他提示词规范文档。" }
      ],
      "used_for": ["确认官方体系内没有第二处会写 object N/background N 的文件"]
    },
    {
      "id": "s09",
      "title": "README.zh-CN.md(存在性与字节数核对)",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/README.zh-CN.md",
      "bytes": 36894,
      "sha256_16": "5c109694dae87cf0",
      "crawl_time": "2026-08-27T06:43:20Z",
      "quotes": [],
      "used_for": ["仅核对存在;逐行差异未比对(knowhow_v3 §7.5)"]
    },
    {
      "id": "s10",
      "title": "openai.yaml(技能 UI 元数据)",
      "url": "https://raw.githubusercontent.com/MiniMax-AI/MiniMax-H3/main/.claude/skills/h3-prompt-writing/agents/openai.yaml",
      "bytes": 240,
      "sha256_16": "7770c9d784b7251a",
      "crawl_time": "2026-08-27T06:44:00Z",
      "quotes": [],
      "used_for": ["确认仅为可选 ChatGPT/Codex UI 元数据,不含规范内容"]
    }
  ],
  "negative_findings": [
    { "claim": "官方原文不存在 object N / background N 字面标签", "checked_by": "grep -i 'object\\|background' 于 s01/s02/s03/s04 全文", "result": "object/background 仅以普通词义出现(objects; Scenes, backgrounds, or environments; background music),无任何编号标签语法" },
    { "claim": "官方文本未出现 references 数组这一请求字段名", "checked_by": "通读 s05/s06 请求体与 s04 相关章节", "result": "参考素材通道名为 conditions[],素材地址在每条 uri 键上" },
    { "claim": "对白标点两条规则表面互斥、作用域不同(新发现)", "checked_by": "s02 L129 与 s03 L274 并排对照", "result": "user-provided 台词须 verbatim 保标点;参考音频直接复用/重演的台词要求标准化标点并删装饰符号——Digester 需拍板作用域条款" },
    { "claim": "人物出场过程 / 道具跨镜跳变 / JSON 序列化通道可解析性均无官方逐字条款", "checked_by": "关键词检索 enters/enter the frame/walks into/consistent/object 连续性 + 通读全部写作规范", "result": "仅能找到相邻规则(s03 L246、s02 L54/L92、保框架 s03 L165-L170),SKILL v0.3.0 中相关条目须标注为内部规范或推演" }
  ],
  "blocked_sources": [
    { "url": "https://platform.minimax.io/docs/api-reference/video-generation-v2-create", "reason": "WebFetch 域安全校验被拒 + curl 超时,三次尝试一致", "impact": "开放平台请求 schema 未核实,勿在本轮报告外推测其字段" },
    { "url": "https://docs.sglang.io/cookbook/diffusion/MiniMax/MiniMax-H3", "reason": "本轮未尝试抓取(README 引用其作为 conditions/media 选项完整清单)", "impact": "role 枚举值可用它做第二信源验证,留作下轮" }
  ]
}