@oh-story/dsh 0.1.14 → 0.1.16

package/lib/drama/skills/short-drama-image-prompts/assets/lookdev-frame-spec.jsonl.md

@@ -1,128 +0,0 @@
-# `lookdev-frame` 填写模板
-
-只在 `purpose: lookdev_frame` 时读取。首行是 `sources` 声明记录，之后每行一个候选风格帧规格；它投影已接受视觉方向，
-不替普通人物、地点、道具规格增加字段。删除当前测试轴不需要的可选项。
-
-## 首行：`sources` 声明
-
-本文件用到的每个上游快照在这里声明一次，键取产物文件名，在本文件内唯一且稳定。视觉方向与
-制作形态同出于 `short-drama.json`，共用一个键：
-
-```json
-{
-  "record_type": "sources",
-  "schema_version": "1.0.0",
-  "sources": {
-    "short-drama": {
-      "owner": "creator",
-      "artifact": "short-drama.json"
-    },
-    "<identity-owner>": {
-      "owner": "short-drama-assets",
-      "artifact": "设定集/<identity-owner-file>.jsonl"
-    },
-    "<variant-owner>": {
-      "owner": "short-drama-assets",
-      "artifact": "设定集/<variant-owner-file>.jsonl"
-    },
-    "screenplay-index": {
-      "owner": "short-drama-write",
-      "artifact": "剧集/<EP>/screenplay-index.jsonl"
-    },
-    "<reference>": {
-      "owner": "<reference-owner>",
-      "artifact": "<project-relative-reference-record>"
-    }
-  }
-}
-```
-
-引用用 `src` 指向声明键，再写本条引用自己的 `record_id` 和 `field`。
-
-## 规格行
-
-```json
-{
-  "spec_id": "LOOKDEV-<stable-id>",
-  "status": "candidate",
-  "purpose": "lookdev_frame",
-  "lookdev_axis": "character_expression | core_location | high_pressure_scene",
-  "direction_ref": {
-    "src": "short-drama",
-    "field": "/creator_authority/visual_direction/choices/look_development"
-  },
-  "production_profile_ref": {
-    "src": "short-drama",
-    "field": "/creator_authority/production_profile"
-  },
-  "subject_bindings": [
-    {
-      "identity_ref": {
-        "src": "<identity-owner>",
-        "record_id": "CHAR/LOC/PROP-<id>"
-      },
-      "variant_ref": {
-        "src": "<variant-owner>",
-        "record_id": "LOOK/VIEW/PSTATE-<id>"
-      },
-      "role": "expression_subject | location | pressure_actor | evidence"
-    }
-  ],
-  "story_context_refs": [
-    {
-      "src": "screenplay-index",
-      "record_id": "BLK-<EP>-<SC>-<kind><nn>",
-      "field": "/<exact-source-field-if-needed>",
-      "role": "scene_heading | action | dialogue | information_permission | story_state"
-    }
-  ],
-  "reference_bindings": [
-    {
-      "slot_id": "REF-<stable-slot>",
-      "order": 1,
-      "artifact_ref": {
-        "src": "<reference>",
-        "record_id": "<stable-record-id>"
-      },
-      "role": "style",
-      "may_control": [
-        "<色彩层级/材质处理/阴影边缘/景深倾向/画面密度>"
-      ],
-      "must_not_control": [
-        "<角色身份/固定地理/剧情状态/人数/持物/道具文字>"
-      ],
-      "admission_status": "unverified | creator_described | visually_inspected",
-      "reference_observation_ref": null,
-      "unresolved_risks": []
-    }
-  ],
-  "test_question": "<本帧要暴露哪一种视觉方向风险>",
-  "stable_visual_rules": [
-    "<跨测试轴必须保留的可观察规则>"
-  ],
-  "allowed_variation": [
-    "<本场允许变化的光比/冷暖/留白/密度>"
-  ],
-  "composition_and_state": "<只写当前代表帧需要的构图、空间与剧情状态>",
-  "appearance": {
-    "materials": [
-      "<材质处理>"
-    ],
-    "palette": "<色彩层级>",
-    "lighting": "<光源逻辑与阴影边缘>",
-    "depth_or_density": "<景深/画面密度>"
-  },
-  "constraints": [
-    "<必须保持的身份、地理、文字与信息权限>"
-  ],
-  "generic_prompt": "<只含可被画出的自然表述，语言按 #/format/prompt_language，不含字段名/hash/审查话术>",
-  "derivation": {
-    "renderer": "generic-markdown"
-  },
-  "provenance": "creator_project"
-}
-```
-
-人物与地点测试没有剧情职责时删除 `story_context_refs` 及其 `sources` 键；高压力场景必须保留真实 `BLK-…`
-来源。单主体也使用一项 `subject_bindings[]`，不再与 `asset_binding` 二选一。无参考媒体时
-`reference_bindings` 为空；多参考的 `slot_id` 稳定、`order` 唯一。

package/lib/drama/skills/short-drama-image-prompts/assets/lookdev-prompts.md

@@ -1,34 +0,0 @@
-# `lookdev-prompts.md` 可复制输出模板
-
-此文件由已接受 Look Development 规格派生。它展示视觉方向怎样跨测试轴保持，
-只复制引用块内的自然语言提示词；本文件不触发图片或媒体服务，实际生产交
-`$short-drama-produce` 并单独确认。
-
-**这是项目级文件，不是某一集的创作文档。** 它不参与 `creator_markdown_check.py` 的
-五文档校验；某一集要用的资产提示词写进该集的《图片提示词.md》，骨架以 `SKILL.md` 为准。
-
-```markdown
-# 项目 Look Development 提示词
-
-> 来源：`lookdev-image-prompt-specs.jsonl` 的已接受记录
-> 视觉方向：`short-drama.json#/creator_authority/visual_direction/choices/look_development`
-> 范围：仅代表帧提示词，不拥有角色身份、地点地理或剧情状态
-
-## `LOOKDEV-<id>` · `<人物表现 | 核心地点 | 高压力场景>`
-
-- **测试问题**：<本帧要暴露的视觉方向风险>
-- **绑定事实**：<准确身份/变体；高压力帧再列 scene/block refs>
-- **跨轴稳定**：<形状、材质、色彩、阴影、景深或密度中的共同规则>
-- **本帧可变**：<当前场景允许变化的部分>
-- **风格参考权限**：`<slot_id / order / reference>` 只决定 `<may_control>`，不得导入 `<must_not_control>`
-- **仍未知**：<只能由授权生产观察回答的风险；无则写“无”>
-
-### 可复制通用提示词
-
-> <只写可被画出的主体、空间、构图、材质、光色和状态；不含字段名、hash、审查话术、供应商参数或生成历史。>
-
----
-```
-
-三类测试帧不要求固定数量或宫格。人物/地点帧没有剧情职责时不列 scene/block refs；高压力帧必须
-显示其准确来源。规格或视觉方向改了就重新派生本文件，不直接手改缓存。

package/lib/drama/skills/short-drama-storyboard/assets/coverage-template.json

@@ -1,49 +0,0 @@
-{
-  "sources": {
-    "shots": {
-      "owner": "short-drama-storyboard",
-      "artifact": "剧集/EP001/storyboard/shots.jsonl"
-    },
-    "screenplay-index": {
-      "owner": "short-drama-write",
-      "artifact": "剧集/EP001/screenplay-index.jsonl"
-    },
-    "short-drama": {
-      "owner": "short-drama",
-      "artifact": "short-drama.json"
-    }
-  },
-  "episode_id": "EP001",
-  "dispositions": [
-    {
-      "block_id": "BLK-EXAMPLE",
-      "status": "covered",
-      "reason": "Primary action and reaction remain continuous.",
-      "shot_refs": [
-        {
-          "src": "shots",
-          "record_id": "SHOT-EXAMPLE",
-          "authority": "candidate"
-        }
-      ]
-    }
-  ],
-  "episode_duration": {
-    "shot_seconds_total": null,
-    "counted_shot_ids": [
-      "SHOT-EXAMPLE"
-    ],
-    "unresolved_durations": [],
-    "target_ref": {
-      "src": "short-drama",
-      "field": "/format/target_seconds_per_episode"
-    },
-    "target_seconds": null,
-    "delta_seconds": null,
-    "disposition": "no_target_declared | within_creator_tolerance | creator_accepted_overrun | to_revise"
-  },
-  "screenplay_ref": {
-    "src": "screenplay-index"
-  },
-  "status": "candidate"
-}

package/lib/drama/skills/short-drama-storyboard/assets/keyframe-prompts.md

@@ -1,36 +0,0 @@
-# `keyframe-prompts.md` 可复制输出模板
-
-这是由已接受的 `keyframes.jsonl` 生成的文本，不是镜头边界或资产事实来源。
-每次导出都记下配方版本；文本被手改时，先预览恢复、
-采用或合并的结果。
-
-**这份模板属于结构化管线**（`keyframes.jsonl` 一路）。creator-first 项目的《分镜.md》骨架
-以 `SKILL.md` 为准：`## SHOT-... · 中文名`，关键帧正文标题是 `### 冻结关键帧提示词`。
-不要把这里的加粗字段名、`KEY-...` 或 `### 可复制通用提示词` 搬进 creator-first 文档。
-
-```markdown
-# EP<编号> · 冻结关键帧提示词
-
-> 来源：`keyframes.jsonl` 的已接受记录
-> 配方：`keyframe-generic@<version>`
-> 范围：单一静止瞬间；不生成图片，不写时间动作
-
-## `SHOT-<id>` · `KEY-<id>`
-
-- **镜头目的**：<观众此刻必须看见什么>
-- **观众可见性**：逐项列出 `<来源 · 展示或暂缓 · 声画载体 · 揭示条件 · 保护方法>`
-- **边界来源**：`SHOT-<id>/start_boundary`
-- **资产绑定**：`<准确人物与造型 · 地点与观察视角 · 道具与状态>`
-- **文字处理**：`<已接受的来源文字政策 -> 本帧呈现方法>`
-
-### 可复制通用提示词
-
-> <焦点主体与识别锚点>。<构图层级、人物/道具/空间锚点关系>。摄影机<景别、角度、镜头意图>。
-> <一个可冻结的身体朝向、姿态、目光、双手、持物与可见表演状态>。<继承的光向、时段、天气与必要气氛>。
-> <文字处理>。保持<边界与身份事实>；排除<本帧高风险漂移>。
-
----
-```
-
-若一句话必须靠“先、再、随后、最终”才能成立，就交给运动提示词环节；若要改变镜头
-开始边界，先向分镜负责人提出修订请求，不要在派生文本中偷偷改。

package/lib/drama/skills/short-drama-storyboard/assets/keyframe-template.jsonl

@@ -1,2 +0,0 @@
-{"record_type":"sources","schema_version":"1.0.0","sources":{"<identity-owner-file>":{"owner":"short-drama-assets","artifact":"设定集/<identity-owner-file>.jsonl"},"<variant-owner-file>":{"owner":"short-drama-assets","artifact":"设定集/<variant-owner-file>.jsonl"},"location-views":{"owner":"short-drama-assets","artifact":"设定集/location-views.jsonl"},"props":{"owner":"short-drama-assets","artifact":"设定集/props.jsonl"},"short-drama":{"owner":"short-drama","artifact":"short-drama.json"},"shots":{"owner":"short-drama-storyboard","artifact":"剧集/<EP>/storyboard/shots.jsonl"}}}
-{"keyframe_id":"KEY-EXAMPLE","status":"candidate","shot_ref":{"src":"shots","record_id":"SHOT-EXAMPLE","authority":"candidate"},"boundary_role":"start | end","boundary_ref":{"src":"shots","record_id":"SHOT-EXAMPLE","field":"/start_boundary | /end_boundary","authority":"candidate"},"purpose":"<frozen viewing purpose>","asset_bindings":[{"identity_ref":{"src":"<identity-owner-file>","record_id":"CHAR/LOC/PROP-<id>"},"variant_ref":{"src":"<variant-owner-file>","record_id":"LOOK/VIEW/PSTATE-<id>"}}],"focal_subjects":[],"composition":{"hierarchy":[],"spatial_relations":[],"foreground":[],"background":[]},"delivery_surface_ref":{"src":"short-drama","field":"/creator_authority/delivery_surface"},"camera":{"framing":"<shot-owned projection>","angle":"<shot-owned projection>","camera_height":"<shot-owned projection>","lens_intent":"<广/标准/长 — frame-only choice; the compression wanted, not a focal length>"},"boundary_projection":{"positions":[],"facing":[],"pose_balance":[],"gaze":[],"hands":[],"held_props":[],"visible_state":[]},"light_projection":{"source_ref":{"src":"location-views","record_id":"VIEW-<id>","field":"/state_differences/light"},"direction":"<direction>","frame_relevant_detail":"<minimal>"},"text_treatment_refs":[{"src":"props","record_id":"PROP-<id>","field":"/text_policy"}],"exclusions":[],"generic_prompt":"<这一瞬间要拍出来的画面：谁在哪、彼此距离、前中后景、视线落点、构图与景别、机位side/俯仰/焦段意图、光从哪来什么质感、每个在场者的姿态朝向手部持物、与上一镜相同的部分写一句「与上一镜相同」、排除项；不写镜头/记录 ID、规则 ID、状态词与成段否定罗列；写满的样子见 references/keyframe-craft.md>","derivation":{"recipe_version":"<version>"}}

package/lib/drama/skills/short-drama-storyboard/assets/shot-template.jsonl

@@ -1,2 +0,0 @@
-{"record_type":"sources","schema_version":"1.0.0","sources":{"<characters-or-props>":{"owner":"short-drama-assets","artifact":"设定集/<characters-or-props>.jsonl"},"<looks-or-prop-states>":{"owner":"short-drama-assets","artifact":"设定集/<looks-or-prop-states>.jsonl"},"location-views":{"owner":"short-drama-assets","artifact":"设定集/location-views.jsonl"},"locations":{"owner":"short-drama-assets","artifact":"设定集/locations.jsonl"},"props":{"owner":"short-drama-assets","artifact":"设定集/props.jsonl"},"screenplay-index":{"owner":"short-drama-write","artifact":"剧集/<EP>/screenplay-index.jsonl"},"short-drama":{"owner":"short-drama","artifact":"short-drama.json"}}}
-{"shot_id":"SHOT-EXAMPLE","status":"candidate","source_refs":[{"src":"screenplay-index","record_id":"BLK-EXAMPLE"}],"coverage_role":"primary | intentional_repeat","purpose":"<what the audience must notice and why this is a new shot>","information_change":"<before -> after for audience/character>","audience_visibility":[{"source_ref":{"src":"screenplay-index","record_id":"BLK-<id>"},"fact":"<本镜必须展示或暂缓的一个事实>","permission":"show_now | withhold_now","carrier":"<可见/可听载体；暂未出现则写 none>","reveal_trigger":"<允许释放的动作/台词/cut；不在本镜则写 future>","protection_method":"<framing/crop/occlusion/offscreen/focus/timing/none，可组合说明>","rationale":"<与 source 信息权限和本镜职责的关系>"}],"location_binding":{"identity_ref":{"src":"locations","record_id":"LOC-<id>"},"variant_ref":{"src":"location-views","record_id":"VIEW-<id>"}},"asset_bindings":[{"identity_ref":{"src":"<characters-or-props>","record_id":"CHAR/PROP-<id>"},"variant_ref":{"src":"<looks-or-prop-states>","record_id":"LOOK/PSTATE-<id>"},"role":"<subject/evidence/background>"}],"duration_seconds":null,"start_boundary":{"positions":["<每条都写成不读上一镜也能理解的绝对事实：谁在哪个锚点的哪一侧；『同上』『位置不变』『同前』要往回翻才解释得通，不算已写>"],"facing":[],"gaze":[],"hands":[],"held_props":[],"visible_state":[]},"primary_transition":{"trigger":"<什么可见或可听事件让动作开始>","action":"<从起点到终点的唯一主要动作；交接时写清接触者、接触方式与落点>","result":"<逐项落到 end_boundary，不在镜外补转换>"},"end_boundary":{"positions":["<同样是绝对事实：本镜结束时人在哪、朝哪，不写成与开始状态的差异>"],"facing":[],"gaze":[],"hands":[],"held_props":[],"visible_state":[]},"framing":{"size":"<extreme_wide 远景/wide 全景/medium 中景/medium_close 中近景/close 近景/close_up 特写/extreme_close_up 大特写/detail 细节镜头 or creator vocabulary>","angle":"<horizontal: 正面/斜侧/正侧/后斜侧/背面 or creator vocabulary; say which side of the axis>","camera_height":"<地平/低(仰)/平视/高(俯)/顶视; 平视 is a choice, not an unset field>","aspect_notes":"<composition constraint>"},"delivery_surface_ref":{"src":"short-drama","field":"/creator_authority/delivery_surface"},"blocking":{"axis":"<working side/bridge/deliberate break>","screen_direction":"<direction>","paths":[],"fixed_anchors":[]},"camera_plan":{"behavior":"locked | pan | tilt | push | pull | track | follow | orbit | crane | handheld | transition","motivation":"<attention/pressure/alignment/reveal/rhythm>","intervals":[]},"dialogue_audio_refs":[],"text_treatment_refs":[{"src":"props","record_id":"PROP-<id>","field":"/text_policy"}],"transition_intent":null,"creator_overrides":[]}

package/lib/drama/skills/short-drama-video-prompts/assets/delivery-container.jsonl.md

@@ -1,101 +0,0 @@
-# `delivery-containers.jsonl` 填写模板
-
-每行一个交付容器对象。容器只是**已接受镜头的排布方式**：它记录哪些镜头按什么顺序装进
-同一个交付单位，以及由此得出的容器时长。它不拥有镜头边界、目的、起止状态或时长——
-这些仍属分镜；容器只引用它们并做加法。
-
-只有创作者声明了分段或多镜打包的交付方式时才需要本文件；单镜交付不必创建容器记录，
-`VID-13` 在那种情形下由运动规格自身的镜头时长满足。
-
-文件第一行是 `sources` 声明：每个上游快照在这里写一次 `owner`、`artifact` 和已接受的
-`artifact`，后面的记录用它的 key 引用。key 用产物文件名派生的短小写名字，在本文件内稳定且唯一。
-
-```jsonl
-{"record_type":"sources","schema_version":"1.0.0","sources":{"shots":{"owner":"short-drama-storyboard","artifact":"剧集/<EP>/storyboard/shots.jsonl"},"motion-specs":{"owner":"short-drama-video-prompts","artifact":"剧集/<EP>/storyboard/motion-specs.jsonl"},"short-drama":{"owner":"creator","artifact":"short-drama.json"}}}
-```
-
-其后每行一个容器记录，引用写 `src` 加它指向的记录 `record_id` 或字段 `field`：
-
-```json
-{
-  "container_id": "CONT-<stable-id>",
-  "status": "candidate",
-  "production_profile_ref": {
-    "src": "short-drama",
-    "field": "/creator_authority/production_profile"
-  },
-  "members": [
-    {
-      "order": 1,
-      "shot_ref": {
-        "src": "shots",
-        "record_id": "SHOT-<id>"
-      },
-      "motion_ref": {
-        "src": "motion-specs",
-        "record_id": "MOTION-<stable-id>"
-      },
-      "accepted_duration_ref": {
-        "src": "shots",
-        "record_id": "SHOT-<id>",
-        "field": "/duration_seconds"
-      },
-      "accepted_duration": "<从 accepted_duration_ref 读到的值，投影不改写>",
-      "location_binding_ref": {
-        "src": "shots",
-        "record_id": "SHOT-<id>",
-        "field": "/location_binding"
-      },
-      "asset_bindings_ref": {
-        "src": "shots",
-        "record_id": "SHOT-<id>",
-        "field": "/asset_bindings"
-      }
-    }
-  ],
-  "container_duration": "<各成员 accepted_duration 之和>",
-  "membership_basis": {
-    "source_order_contiguous": "<true | 说明为什么不连续及去向>",
-    "binding_chain_equal": "<true | 说明哪一位成员的绑定不同、以及为什么仍同容器>",
-    "scene_boundary_not_crossed": "<true | 说明跨越原因与已接受依据>"
-  },
-  "unresolved": [],
-  "provenance": "creator_project"
-}
-```
-
-## 结构校验点
-
-`VID-13` 靠这条记录本地可证，不依赖阅读渲染文本。带 ✓ 的由 [container_check.py](../scripts/container_check.py) 判定，其余由审查者判定——此前这份清单整体写成「本地可证」，而脚本只实现了其中两项：
-
-1. ✓ `members[]` 非空，`order` 唯一、连续、升序；
-2. 每条引用的 `src` 在本文件 `sources` 声明里找得到对应条目；
-3. ✓ 每个成员的 `accepted_duration` 等于其 `accepted_duration_ref` 解析后指向的分镜值；
-4. ✓ `container_duration` 等于各成员 `accepted_duration` 之和；
-5. 每个成员的 `motion_ref` 指向的运动规格，其 `shot_ref` 与该成员 `shot_ref` 解析到同一镜头记录；
-6. **逐成员**解析 `location_binding_ref` 与 `asset_bindings_ref`，各成员解析结果相同时
-   `binding_chain_equal` 才能为 `true`；只引用其中一条成员记录不构成证明。
-7. ✓ `membership_basis` 三项都有结论，未成立的写进 `unresolved`，不留空。
-
-任何一项不成立即为结构缺陷，按主技能的 `stale` 与恢复流程处理，不在渲染文本里补救。
-
-## 依赖方向是单向的
-
-容器 → 运动规格 → 镜头，**不存在反向的文件引用**。运动规格不带 `container_ref`：
-两端在各自 `sources` 里互相声明对方会形成环：依赖方向必须单向，否则谁都不能先落盘。
-文字上写“只读”不能消除这个环。
-
-要从一个镜头反查它属于哪个容器，扫描容器记录的 `members[]`，不在运动规格里存副本。
-
-## 与其他记录的关系
-
-- **运动规格**：每条运动规格仍然只绑定一个 `shot_ref`，且不感知容器的存在。
-- **渲染文本**：`video-prompts.md` 的容器一节由本记录派生，是缓存，不是权威；文本与本
-  记录不一致时以本记录为准。
-- **成员时长**：一律是分镜的只读投影。要改时长就改分镜，然后重算 `container_duration`；
-  不得在容器里直接改数。
-- **成员资格判据**：语义部分见 `references/delivery-profile.md` 的多镜容器成员资格；
-  本文件只负责让判据的结论可被引用与核对。
-
-复制后删除不适用的可选字段。容器不跨越已接受的场次或时间跳跃；省略、闪回分支与声明过的
-蒙太奇各自单独成容器。

package/lib/drama/skills/short-drama-video-prompts/assets/motion-spec.jsonl.md

@@ -1,217 +0,0 @@
-# `motion-specs.jsonl` 填写模板
-
-文件第一行是 `sources` 声明：每个上游快照在这里写一次 `owner`、`artifact` 和已接受的
-`artifact`，后面的记录用它的 key 引用。key 用产物文件名派生的短小写名字，在本文件内稳定且唯一。
-本阶段插入的片段（[`performance.fragment.json`](performance.fragment.json)、
-[`coverage-scope.fragment.json`](coverage-scope.fragment.json)）落进同一个文件，它们用到的 key
-也在这里声明。
-
-```jsonl
-{"record_type":"sources","schema_version":"1.0.0","sources":{"shots":{"owner":"short-drama-storyboard","artifact":"剧集/<EP>/storyboard/shots.jsonl"},"keyframes":{"owner":"short-drama-storyboard","artifact":"剧集/<EP>/storyboard/keyframes.jsonl"},"screenplay-index":{"owner":"short-drama-write","artifact":"剧集/<EP>/screenplay-index.jsonl"},"characters":{"owner":"short-drama-assets","artifact":"设定集/characters.jsonl"},"short-drama":{"owner":"creator","artifact":"short-drama.json"}}}
-```
-
-其后每行一个候选运动规格对象。引用写 `src` 加它指向的记录 `record_id` 或字段 `field`；
-`boundary_refs` 只读，候选状态属于本运动规格，不向已经接受的上游引用传播。不要加入
-`duration_override`、`end_override` 或 `next_shot_write`。以下字符串只说明怎样填写，
-不是固定答案。
-
-```json
-{
-  "motion_id": "MOTION-<stable-id>",
-  "status": "candidate",
-  "shot_ref": {
-    "src": "shots",
-    "record_id": "SHOT-<id>"
-  },
-  "keyframe_ref": {
-    "src": "keyframes",
-    "record_id": "KEY-<id>"
-  },
-  "production_profile_ref": {
-    "src": "short-drama",
-    "field": "/creator_authority/production_profile"
-  },
-  "boundary_refs": {
-    "duration": {
-      "src": "shots",
-      "record_id": "SHOT-<id>",
-      "field": "/duration_seconds",
-      "value_seconds": 0.0
-    },
-    "start": {
-      "src": "shots",
-      "record_id": "SHOT-<id>",
-      "field": "/start_boundary"
-    },
-    "primary_transition": {
-      "src": "shots",
-      "record_id": "SHOT-<id>",
-      "field": "/primary_transition"
-    },
-    "end": {
-      "src": "shots",
-      "record_id": "SHOT-<id>",
-      "field": "/end_boundary"
-    },
-    "next_start": {
-      "src": "shots",
-      "record_id": "SHOT-<next-id>",
-      "field": "/start_boundary",
-      "access": "comparison_only"
-    }
-  },
-  "reference_bindings": [
-    {
-      "slot_id": "REF-<stable-slot>",
-      "order": 1,
-      "artifact_ref": {
-        "src": "keyframes",
-        "record_id": "KEY-<id>"
-      },
-      "role": "start_frame | continuity_video | actual_tail_frame",
-      "may_control": [
-        "<本镜 accepted 起始构图与可见状态>"
-      ],
-      "must_not_control": [
-        "<尚未发生的动作/终态/无权威文字>"
-      ],
-      "admission_status": "unverified | creator_described | visually_inspected",
-      "reference_observation_ref": null,
-      "unresolved_risks": [
-        "<没有观察证据时保留的文字/水印/裁切风险>"
-      ]
-    }
-  ],
-  "start_anchor": {
-    "pose_balance": "<仅运动必需>",
-    "gaze": "<目标>",
-    "hands": {
-      "left": "<状态>",
-      "right": "<状态>"
-    },
-    "held_props": [
-      "<exact binding + hand>"
-    ],
-    "spatial_relations": [
-      "<与行动对象的关系>"
-    ]
-  },
-  "ordered_subject_motion": [
-    {
-      "order": 1,
-      "actor": "<asset binding>",
-      "trigger": "<accepted cue>",
-      "action": "<可见动作>",
-      "direction_or_path": "<方向/路径>",
-      "object_or_contact": "<对象/接触>",
-      "result": "<阶段结果>",
-      "timing": {
-        "mode": "relative | explicit",
-        "value": "<顺序词或秒区间>"
-      }
-    }
-  ],
-  "camera": {
-    "behavior": "locked | move | transition",
-    "motivation": "reveal | pressure | alignment | relationship | transition | deliberate_stillness",
-    "intervals": [
-      {
-        "range": "<相对阶段或秒区间>",
-        "mode": "<lock/pan/tilt/dolly/handheld/follow>",
-        "path_tempo": "<方向/节奏>",
-        "endpoint": "<在 accepted framing/boundary 内>"
-      }
-    ]
-  },
-  "environment_motion": [
-    {
-      "element": "<已有环境元素>",
-      "motion": "<有剧情意义的变化>",
-      "cause": "<连续性/主体动作>"
-    }
-  ],
-  "audio": [
-    {
-      "source_ref": {
-        "src": "screenplay-index",
-        "record_id": "BLK-<EP>-<SC>-D<nn>"
-      },
-      "speaker_ref": {
-        "src": "characters",
-        "record_id": "CHAR-<id>"
-      },
-      "voice_direction_ref": {
-        "src": "characters",
-        "record_id": "CHAR-<id>",
-        "field": "/voice_direction"
-      },
-      "kind": "dialogue | VO | OS | SFX | ambience | music",
-      "exact_text": "<仅 source 有文本时逐字引用>",
-      "delivery_or_spatial_intent": "<不改文本的表演/声源/层级>",
-      "timing": "<相对阶段或秒区间>"
-    }
-  ],
-  "timing_plan": {
-    "mode": "relative | explicit",
-    "phases": [
-      "<阶段、overlap 与 landing 空间>"
-    ],
-    "declares_overlap": false,
-    "declared_total_or_endpoint_seconds": 0.0
-  },
-  "end_report": {
-    "projection": {
-      "pose": "<reported>",
-      "position": "<reported>",
-      "gaze": "<reported>",
-      "hands": "<reported>",
-      "held_props": "<reported>",
-      "visible_state": "<reported>"
-    },
-    "comparison": "match | mismatch | unrealized",
-    "differences": []
-  },
-  "reference_frame_economy": {
-    "frame_carries": [
-      "appearance",
-      "composition",
-      "base lighting"
-    ],
-    "repeated_for_motion_only": [
-      "<hand/prop/path 等必要局部>"
-    ]
-  },
-  "creator_overrides": [
-    {
-      "rule_id": "<VID-*>",
-      "choice": "<覆盖>",
-      "rationale": "<理由>"
-    }
-  ],
-  "generic_prompt": "<把本规格渲染成一段只含要拍出来的画面的交付文本：从已接受起点的姿态与持物说起，逐条写动作、接触、摄影机行为与终点状态；不写镜头/记录 ID、规则 ID、状态词、工艺备注与成段否定罗列；写满的样子见 references/production-prompt-grammar.md>",
-  "derivation": {
-    "recipe_version": "<version>"
-  },
-  "provenance": "creator_project"
-}
-```
-
-
-运动规格**不带指回交付容器的引用**。依赖方向只有一条：容器 → 运动规格 → 镜头。两端在各自
-`sources` 里互相声明对方会形成环——依赖方向必须单向，
-永远得不到可发布的稳定快照。要找某个镜头属于哪个容器，从容器记录的 `members[]` 反查，不在本文件里
-存副本。容器记录见 [delivery-container.jsonl.md](delivery-container.jsonl.md)。
-
-默认 master 不重复 `purpose_ref` 或 `coverage_scope`：镜头目的、场次计划与原文覆盖从准确
-`shot_ref` 及其上游读取。只有 pickup/alternate 才按
-[motion-recipe.md](../references/motion-recipe.md) 增加 `coverage_scope`，记录相对母版的补充、保留与去向。
-`boundary_refs` 只保留结构校验需要的 duration/start/end/next-start 精确字段投影，不再重复镜头目的、
-场次计划或原文职责；删除它前必须先让 timing 与 continuity 校验器能从 `shot_ref` 安全解析同一快照。
-
-普通记录省略 `performance_arcs[]` 与 `attention_handoffs[]`；存在可见表演变化或注意交接时，按
-[`performance.fragment.json`](performance.fragment.json) 插入完整字段。空镜、道具细节、纯空间
-转场和只有物理动作的镜头不编造 arc。多参考的 `slot_id` 稳定且 `order` 唯一。
-`reported_end` 只作比较；末镜没有真实下一镜时改用
-`next_start_locator`。附加参考为空时使用空数组；对白说话者与声音方向只有在已接受引用存在时才填写。
-母版、补拍和替代关系保留在同一规格文件内，替代决定由审查结论拥有。具体取舍按
-`references/motion-recipe.md` 与 `references/review-and-fixtures.md` 判断。

package/lib/drama/skills/short-drama-video-prompts/assets/music-spec.jsonl.md

@@ -1,31 +0,0 @@
-# Timeline Music Spec
-
-Music is a timeline layer, not a default bed baked into every generated shot.
-Create a record only when accepted story or directing material gives music a
-specific dramatic job.
-
-The file opens with a `sources` header that declares each upstream snapshot
-once. A source key is short, lowercase and derived from the artifact basename;
-it stays stable and unique within the file.
-
-```jsonl
-{"record_type":"sources","schema_version":"1.0.0","sources":{"screenplay":{"owner":"short-drama-write","artifact":"剧集/EP001/screenplay.md"}}}
-```
-
-Every record after the header names the snapshot through its key:
-
-```json
-{"music_id":"MUS-EP001-01","scope":{"episode_id":"EP001","start_seconds":0,"end_seconds":12},"source_refs":[{"src":"screenplay","record_id":"EP001-SC001"}],"narrative_function":"<what changes for the audience, and why picture/performance alone is insufficient>","prompt":"<provider-neutral style, mood, instrumentation, energy and scene description>","mode":"instrumental | song","lyrics":null,"mix_intent":{"entry":"<relative entry against picture or dialogue>","exit":"<relative exit or transition>","duck_under_dialogue":true,"loop":false},"status":"candidate"}
-```
-
-- `source_refs` carry `src` plus the record or field they point at; the
-  artifact's `owner` and `artifact` live in the header entry that `src`
-  names.
-- `scope` is the accepted timeline interval; it does not promise a supplier can
-  generate that exact duration.
-- `prompt` contains musical intent only, never model names, API fields, artist
-  imitation, remote task identifiers or credentials.
-- `song` requires creator-owned or licensed `lyrics`; `instrumental` carries no
-  lyrics. Do not ask a provider to silently rewrite accepted lyrics.
-- `mix_intent` is an editorial handoff. Production generates a source track;
-  editing still owns exact fades, ducking and placement.