@oh-story/dsh 0.1.15 → 0.1.17

package/lib/drama/skills/short-drama-produce/references/adapter-contract.md

@@ -1,5 +1,10 @@
 # Adapter Contract
 
+The bundled provider compilers accept video-reference roles, but the current creator-first Markdown
+source selector declares only image references and character voice references. It cannot bind a previous
+video for continuation. Keep that route unsubmitted until the source interface supports both actual
+video and tail-frame inputs; do not omit the source selector to bypass this boundary.
+
 ## Job file
 
 ```json
@@ -37,12 +42,12 @@
   canonical `剧集|episodes/<EP>/` creator document, matched to that document:
   `图片提示词.md` takes `IMG-*` (image), `视频提示词.md` takes `MOTION-*` (video),
   and `分镜.md` takes `SHOT-*` (image, whose prompt is the shot's
-  `### 冻结关键帧提示词` body rather than `### 可复制提示词`). A new image/video job pointing to either canonical filename must
+  `### 冻结关键帧提示词` body rather than `### 可复制提示词`). A new image/video job using one of these canonical sources must
   provide the matching selector; an arbitrary Markdown file cannot impersonate a
   creator source. `prepare` selects that section and requires `prompt` to exactly
   equal its copyable prompt after Markdown quote markers are removed.
 - `reference_bindings`: zero to sixteen ordered semantic bindings. Every entry has
-  exactly `slot_id`, contiguous `order`, project-relative `path`, Chinese `label`,
+  `slot_id`, contiguous `order`, project-relative `path`, Chinese `label`,
   non-empty `role`, and non-empty `may_control` / `must_not_control` lists. When
   `source_entry` is present, these fields must exactly match that entry's
   `参考` (IMG) or `输入参考图` (SHOT/MOTION) declaration (except `role`, which is
@@ -54,9 +59,9 @@
   the creator document, not this schema, is the authority on. This schema keeps
   `role` free-form because an external adapter names its own inputs, but the
   bundled video adapters read it as the provider's own role and accept only
-  their published vocabulary: `first_frame`, `last_frame`, `reference_image`,
-  `reference_video`, `reference_audio` for MiniMax, and the three `reference_*`
-  values for Seedance. A video job carrying references without bindings fails
+  their implemented vocabulary: `first_frame`, `last_frame`, `reference_image`,
+  `reference_video`, and `reference_audio` for both MiniMax and Seedance.
+  Supported combinations still depend on the configured model. A video job carrying references without bindings fails
   closed there rather than having a role guessed for it.
   A `MOTION-*` entry's `参考音频` line binds after the pictures: its slots follow
   the `输入参考图` bindings in audio `顺序`, continuing the job's `order`, each with
@@ -64,6 +69,17 @@
   slots may use that role. Leaving a declared voice or its character out, or swapping
   characters, fails `prepare`. Adapters must preserve this character binding; bundled
   compilers derive the voice label from `character`, not the free-form asset `label`.
+  For a job rendering a SHOT's own start frame, `prepare` may omit the creator-supplied
+  `PLAN-...` slots because those external pictures are not this job's inputs;
+  the declared real `REF-...` files must still match, and their remaining order must be
+  contiguous from 1. If skipping PLAN leaves gaps, the storyboard owner first revises
+  the source declaration; renumbering the job alone fails source matching.
+  Binding paths must also be unique. Two creative slots may describe different uses of
+  one picture, but this interface cannot submit both with the same path. Have the owning
+  skill and creator decide whether to revise the reference duties or wait for interface
+  support; do not silently merge or omit slots in the job.
+  A job that needs to send the PLAN
+  pictures instead must first bind actual project files as REF slots.
 - `references`: zero to sixteen current project files actually sent to production.
   It may be omitted when `reference_bindings` is present, in which case the paths
   are derived in binding order. If both are present, they must match exactly.
@@ -79,11 +95,7 @@
 
 `prepare` records internal digests of source/reference bytes and returns the exact confirmation phrase. Callers never
 calculate those digests. A changed input requires prepare and confirmation again.
-Jobs already prepared and stored before this contract remain readable under their
-original fingerprint. New non-creator structured jobs remain valid without
-`source_entry` and `reference_bindings`; they still receive the existing
-path/digest/confirmation checks. A `music` job may continue to use the timeline
-music section of canonical `视频提示词.md` without pretending that section is a
+A `music` job uses the timeline music section of canonical `视频提示词.md` without pretending that section is a
 `MOTION-*` entry; other new image/video jobs using canonical creator paths must use
 their matching selector.
 
@@ -106,10 +118,10 @@ Keep this file outside the project:
 let the adapter read its environment or operating-system credential store.
 
 A profile may add `"reference_roles": [...]`, the roles its configured model accepts. When present,
-`run` refuses any binding whose role is not listed. A `reference_audio` binding always needs it:
-a model without an audio input can drop the voice without an error. Either way `run` stops before
-the confirmation is spent or the adapter starts. List `reference_audio` only for a model you know
-accepts reference audio.
+`run` refuses any binding whose role is not listed. A `reference_audio` binding always requires
+that declaration, so an unsupported model cannot silently drop a voice. A role rejection happens
+before confirmation is consumed or the adapter starts. List `reference_audio` only for a model
+known to accept reference audio.
 
 ## Capability sources
 
@@ -123,21 +135,22 @@ The adapter receives one UTF-8 JSON document as raw stdin bytes. Read the
 binary stream (for example `json.load(sys.stdin.buffer)`) rather than relying
 on the machine locale. The document contains the confirmed job plus:
 
-- `run_id`: unique attempt ID;
-- `project_root`: local absolute path to a private, run-scoped snapshot containing
+- `run_id`: the original attempt's unique ID, also reused when collecting that attempt;
+- `project_root`: on a run call, the local absolute path to a private, run-scoped snapshot containing
   the exact confirmed `source` and `references`. It is not the live creator
   project and is deleted after the attempt.
 - `output_root`: empty private, run-scoped staging directory owned by the
   production tool. Every adapter output source must be a direct regular-file
   child of this directory. The whole directory is deleted after success or
   failure.
-- `handle_path`: absolute path where the adapter records the provider's task id
+- `handle_path`: on a run call, the absolute path where the adapter records the provider's task id
   **as soon as it has one, before it starts polling** — write
   `{"provider_job_id": "<id>"}`, atomically. Unlike `output_root` this path
   survives the attempt.
 - `collect_provider_job_id`: present only on a collect call. When it is set the
   adapter must **not submit anything**; it polls and downloads that existing
-  task and returns its outputs as usual.
+  task and returns its outputs as usual. A collect call has no input snapshot or
+  `project_root`; it must not reopen the source or reference files.
 
 ## Recovering a submitted task
 

package/lib/drama/skills/short-drama-produce/references/providers/gpt-image-2.md

@@ -9,7 +9,7 @@ Adapter command:
 Required environment: `OPENAI_API_KEY`. `OPENAI_BASE_URL` optionally overrides the default
 `https://api.openai.com/v1` and must remain HTTPS.
 
-The job must have modality `image` and exactly one output. With no references, the adapter sends JSON to
+The job must use modality `image` and have exactly one output. With no references, the adapter sends JSON to
 `POST /images/generations`. With one to sixteen references, it sends an `image[]` multipart edit to
 `POST /images/edits`. The model is always `gpt-image-2`, `n` is always one, and the single returned `b64_json` image is
 written to a private temporary file.

package/lib/drama/skills/short-drama-produce/references/providers/minimax-h3-video.md

@@ -50,13 +50,18 @@ frame conditioning (`first_frame` / `last_frame`) and full-reference conditionin
 (`reference_image` / `reference_video` / `reference_audio`) are mutually exclusive in one request.
 For continuation, bind the previous actual video as `reference_video` and its actual tail as
 `reference_image`; the compiler labels them `<Video 1>` and `<Picture 1>` in its appended contract.
-Never relabel that tail as `first_frame` while a reference video is present. Reference URLs may be HTTPS, `mm_file://{file_id}`, or a `data:<mime>;base64,<...>` URI — all three
-are documented inputs. A project-relative reference in the confirmed job is read from disk and sent
+This describes compiler support. The current creator-first source selector cannot declare that video,
+so continuation remains unsubmitted in the bundled workflow; see the [source-interface boundary](../adapter-contract.md).
+Never relabel that tail as `first_frame` while a reference video is present. The compiler accepts HTTPS, `mm_file://{file_id}`, or `data:<mime>;base64,<...>` reference URLs.
+The bundled runtime reads each confirmed project-relative reference from disk and sends it
 inline as a data URI, so binding an image in the creator documents is enough to run the job; no
 upload service is required. Each file's bytes must match the media type its extension claims, and
 the published per-modality caps apply: 30MB per image, 50MB per video, 15MB per audio clip, and 64MB
-for the whole request measured after base64 expansion. A reference past those caps fails closed with
-an explicit message; host it and bind an HTTPS URL instead of splitting it.
+for the provider request. The local guard checks individual files and the combined base64 reference
+strings against these caps; other request fields also contribute to the provider's body limit.
+An over-limit reference fails before submission. This production tool accepts project files,
+so a hosted URL cannot replace its `references` path. Have the responsible skill prepare a suitable
+input, preserve the accepted content and control scopes, then preview and confirm the updated job.
 
 Each reference takes its provider role from the confirmed job's `reference_bindings[].role`, which is
 where the creator document's `用途` is translated for this provider. A job that carries references
@@ -71,8 +76,8 @@ status fails closed.
 This release generates audio in the same pass as the picture. Its direct prompt dialect is also
 structured: base/first-frame/first-last-frame tasks use `integrated_multimodal_description`,
 `overall_soundscape`, and `non_diegetic_music`; full-reference tasks use the six-section H3 reference
-form. Chinese dialogue remains exact inside `<d>[Chinese] ...</d>`. Full details live in the
-video-prompt skill's `references/minimax-h3.md`.
+form. Chinese dialogue remains exact inside `<d>[Chinese] ...</d>`. Use
+`$short-drama-video-prompts` for the full H3 prompt dialect.
 
 State the intended sound layers in the prompt. For no non-diegetic music, the
 H3 dialect uses `non_diegetic_music: N/A`; small probes observed unwanted music

package/lib/drama/skills/short-drama-produce/references/providers/minimax-music.md

@@ -11,11 +11,11 @@ Required environment: `MINIMAX_API_KEY`. `MINIMAX_BASE_URL` optionally overrides
 
 The job uses the suite's `music` modality and must have exactly one output. The production prompt becomes the
 music style/mood prompt. Supported public parameters are `lyrics`, `is_instrumental`, `sample_rate`,
-`bitrate`, and `format`. Vocal jobs require exact accepted lyrics. The requested format must match the target extension
+`bitrate`, and `format`. The requested format must match the target extension
 and be `mp3` or `wav`.
 
 For this confirmed-production profile, `lyrics_optimizer` is rejected even when `true` is requested: supplier-authored lyrics would not have
-appeared in the preview and cannot satisfy creator ownership or licensing. A vocal job therefore requires the exact
+appeared in the preview. A vocal job therefore requires the exact
 accepted lyrics; an instrumental job carries neither lyrics nor an optimizer request.
 
 The request always uses model `music-3.0`, non-streaming `output_format: hex`, and

package/lib/drama/skills/short-drama-produce/references/providers/minimax-speech.md

@@ -21,11 +21,13 @@ endpoint, exactly as for the video providers. The voice is a creative decision r
 ## Voice selection
 
 Read the provider's voice listing and confirm availability for the chosen model and account.
-The adapter validates the ID format, not catalogue membership. Record the selected ID in the
-character's 声音方向 and pass it through the confirmed job.
+The adapter validates the ID format, not catalogue membership. Have `$short-drama-assets` record
+the selected ID in the character's 声音方向, then pass it through the confirmed job.
 
 This workflow selects preset voices. The adapter does not enrol or clone a voice;
-any supplied recording retains its documented authorization and usage scope.
+any supplied recording retains its documented authorization and usage scope. This adapter does
+not send that recording as a voice reference. Keep performance instructions outside the spoken
+prompt and express them through the supported parameters.
 
 ## Request shape
 
@@ -35,7 +37,7 @@ The request is non-streaming with `output_format: hex`, `POST /t2a_v2`, and carr
 `data.audio` as hexadecimal bytes, and writes it to a private temporary file.
 
 `emotion` accepts the provider's seven values: `happy`, `sad`, `angry`, `fearful`, `disgusted`,
-`surprised`, `neutral`. Anything outside that set is refused rather than passed through — a rejected
-emotion is a typo caught before it is paid for.
+`surprised`, `neutral`. Anything outside that set is refused before submission; it may be an
+unsupported direction rather than a typo and needs an explicit mapping to an available value.
 
 Protocol reference: [MiniMax Text to Speech](https://platform.minimax.io/docs/api-reference/speech-t2a-http).

package/lib/drama/skills/short-drama-produce/references/providers/seedance.md

@@ -36,23 +36,29 @@ lengthen a short job. Fixed-duration delivery needs an explicit accepted duratio
 output needs its actual media length and speech checked before timing is accepted. The API response's
 integer `duration` rounds down the frame-based duration, so it is not a precise media measurement.
 
-`compile_seedance_payload` accepts explicit `reference_image`, `reference_video`, and
-`reference_audio` roles with matching HTTPS, `asset://` or base64 data URI values. It emits the matching multimodal
+`compile_seedance_payload` accepts explicit `first_frame`, `last_frame`, `reference_image`,
+`reference_video`, and `reference_audio` roles with matching HTTPS, `asset://` or base64 data URI values.
+Each frame role may appear once; supported role combinations depend on the configured model.
+It emits the matching multimodal
 content type and uses `@图片N`、`@视频N`、`@音频N` in its appended reference contract. The bundled
 runtime sends a project-relative reference inline as a `data:<mime>;base64,<...>` URI, which the
 official contract lists alongside a public URL and `asset://`, so binding an image in the creator
 documents is enough to run the job. Each file's bytes must match the media type its extension claims.
-Volcengine publishes no per-file size numbers, so this adapter applies the same conservative guard as
-the MiniMax path — 30MB per image, 50MB per video, 15MB per audio clip, 64MB per request after base64
-expansion — and fails closed above them rather than sending a body the provider may reject. Each
+This adapter applies the same local guards as the MiniMax path: 30MB per image, 50MB per video,
+15MB per audio clip, and 64MB for combined base64 reference strings. These guards do not establish
+the configured model's native limits; request metadata adds further bytes. An over-limit input
+fails before submission. Each
 reference takes its provider role from the confirmed job's `reference_bindings[].role`; a job with
 references but no bindings fails closed rather than guessing.
 
 Prompt wording for Seedance 2.0 is version-specific: normal reference uses “参考 @视频1”, while
-continuation says “向后延长 @视频1” and binds the actual tail as `@图片1`. See the video-prompt
-skill's `references/seedance-2.0.md`.
+continuation says “向后延长 @视频1” and binds the actual tail as `@图片1`. Use
+`$short-drama-video-prompts` for the version's prompt dialect.
+These are compiler capabilities; the current creator-first source selector cannot declare the
+continuation video, so that route remains unsubmitted in the bundled workflow. See the
+[source-interface boundary](../adapter-contract.md).
 
-Seedance 2.5 has a separate `references/seedance-2.5.md` dialect and may carry
+Seedance 2.5 has its own prompt dialect and may carry
 `omni_reference_task_type: reference|edit|extend|auto`. The compiler enforces the documented hard
 parameter combinations for `edit` (`adaptive`, `-1`, reference video) and `extend` (`adaptive`,
 reference video). It does not infer the task type from prompt keywords, because the API checks that

package/lib/drama/skills/short-drama-review/SKILL.md

@@ -1,13 +1,13 @@
 ---
 name: short-drama-review
-description: 审查短剧项目中的原著分析、故事、剧本、视觉设定、图片提示词、分镜、冻结关键帧、视频提示词和已有媒体。用户提出“审稿/检查剧本”“检查资产或连续性”“检查图片/视频提示词”“检查原著分析”“审查模板感”“根据生产观察做项目校准”时使用；只写审查问题、结论和修订要求，不代替 owner 修改来源文件。
+description: 审查短剧项目中的原著分析、故事、剧本、视觉设定、图片提示词、分镜、冻结关键帧、视频提示词和已有媒体。用户提出“审稿/检查剧本”“检查资产或连续性”“检查图片/视频提示词”“检查原著分析”“审查模板感”“根据生产观察做项目校准”时使用；只写审查问题、结论和修订要求，不代替负责人修改来源文件。
 license: MIT
 ---
 
 # 短剧审查
 
-优先由未参与当前版本创作的 reviewer 执行；条件不允许时可以自检并如实说明。审查与修改是两个工作
-单元：本技能只定位问题、影响和必须达到的修订结果，不在同一轮替 owner 改来源。
+优先由未参与当前版本创作的 reviewer 执行；条件不允许时可以自检并如实说明。本技能定位问题、影响和
+必须达到的修订结果，将修改交回负责技能；审查与修改分开执行。
 
 ## Quick Start
 
@@ -28,7 +28,7 @@ license: MIT
 - 证据：……
 - 影响：……
 - 修订结果：由 short-drama-storyboard 补入准确文字及可读条件。
-- 规则：SHT-01 · reviewed_invariant
+- 规则：SHT-01 · structural_invariant
 ```
 
 引用只使用文件名、标题 ID、行号或短引文；不建立 sources 声明、哈希、record ID 或第二套状态。
@@ -60,7 +60,7 @@ license: MIT
 
 ## 工作流
 
-### 1. 冻结范围
+### 1. 确定范围
 
 写清当前要审哪些文件、章节、镜头或提示词，以及创作者已经明确的限制。目标在审查中发生变化时，
 重新读取再下结论。
@@ -79,8 +79,8 @@ license: MIT
 
 ### 3. 带证据审内容
 
-每个 finding 包含：位置、必要短引文或冲突事实、观众/制作影响、必须达到的修订结果、owner、严重
-程度和规则等级。不能只说“AI 味”“不够电影感”或给无证据分数。
+每个 finding 包含位置、必要短引文或冲突事实、观众/制作影响、必须达到的修订结果、负责技能、严重
+程度、规则等级和 `open` / `closed` 状态。“AI 味”“不够电影感”或无证据分数不能替代这些内容。
 
 ### 4. 跨文档综合
 
@@ -94,10 +94,10 @@ license: MIT
 
 - `APPROVE`：没有阻断问题；
 - `APPROVE_WITH_NOTES`：只有不阻断的改进；
-- `REVISE`：存在结构、内容或限制冲突；
+- `REVISE`：存在阻断交付的结构缺陷、经证据确认的内容缺陷或已接受限制冲突；
 - `PROVISIONAL`：关键输入不足，暂时无法完成判断。
 
-按 owner 分组修订要求。本轮完成后交还控制权；修改和复审只有用户明确请求时开始。
+按负责技能分组修订要求。本轮完成后交还控制权；修改和复审只有用户明确请求时开始。
 
 ## 严重程度
 

package/lib/drama/skills/short-drama-review/agents/openai.yaml

@@ -1,4 +1,4 @@
 interface:
   display_name: "短剧审查"
   short_description: "审查短剧故事、资产、连续性、分镜与提示词并给出证据"
-  default_prompt: "使用 $short-drama-review 对短剧项目产物做独立、带证据的校验与审查。"
+  default_prompt: "使用 $short-drama-review 审查我点名的短剧项目产物，给出位置明确、有证据的结论；独立 reviewer 不可用时如实标注自检。"

package/lib/drama/skills/short-drama-review/references/anti-template-repair.md

@@ -1,6 +1,6 @@
 # 去模板感：诊断、修订与误报
 
-“像模板”不是稳定 finding。先定位重复机制如何伤害当前故事，再把修订交给 owner。
+先定位重复机制怎样伤害当前故事，再把具体修订目标交给负责技能；“像模板”本身不足以构成 finding。
 
 ## 目录
 
@@ -20,21 +20,20 @@
 
 只有能引用至少两个位置并说明损失，才发重复 finding。单次使用类型惯例不是模板罪证。
 
-表达层还要检查三种不靠逐字重复也会产生模板感的情况：
+表达层还要检查不依赖逐字重复的情况：
 
 - **过度解释**：动作、证据或台词已经表达人物选择，后面又补一句意义说明、心理翻译或
   道德总结，夺走观众推断空间；
 - **过度收口**：每场都用总结句、金句或完整结论封死余波，下一场只能重新制造压力；
 - **过度工整**：每轮台词都配相同长度的微表情/手部动作，每场都以同构对仗或反问落点，
   人物和场面失去自己的节奏。
-- **信息密度均匀**：每一拍携带的新信息量几乎相同，于是没有一件事像"事件"。这与过度
-  工整不同——那讲的是形式对称，这讲的是**疏密**：一场戏里本该有几拍是白给的、有一拍
-  是砸下来的。单看每一拍都合格，所以逐拍检查抓不到它；要把一场戏的各拍并排看信息量。
-- **代价即时结清**：本集造成的代价全部在本集内还完，于是什么都不积累。这是过度收口的
-  跨集版本。查法很具体——把本集造成的代价与它实际落地的集号列出来，若全部落在同一集，
-  就是这个病；短剧靠未结清的债务把观众带到下一集。
-
-这不是禁词表。Reviewer 仍要引用当前作品中至少两个位置，说明被重复解释、封死或抹平的
+- **信息密度均匀**：每拍携带的新信息量几乎相同，重要事件没有获得相应分量。把一场戏的
+  各拍放在一起比较，检查信息接收、等待和关键变化是否形成疏密；逐拍成立仍可能掩盖整场节奏单一。
+- **代价被自动清空**：连载已经把某项代价设为持续压力，却在成功后自动复位，后续不再继承。
+  对照代价建立的位置与后续落点，说明丢失了什么后果。代价在本集得到有因果的解决、单集完整收束，
+  或创作者有意选择开放与封闭的结尾，本身都不构成缺陷。
+
+Reviewer 要引用当前作品中至少两个位置，说明被重复解释、封死或抹平的
 具体意义；有意的仪式、喜剧节拍、人物口癖、方言和创作者选择继续按下文反例处理。
 
 ## 2. 修订不是换同义词
@@ -85,7 +84,7 @@ Finding 应指出缺少哪个 accepted identity/scale/material/state/text policy
 - 人物因创伤反复做同一动作，作品正在表现停滞而非遗漏变化。
 
 Reviewer 必须说明：重复是否是 creator choice、第二次新增了什么、若保留会伤害什么。
-不确定时记 note/question，不是 REVISE。
+无法确认重复造成损失时，记为 note/question，保留创作者选择。
 
 ## 4. Finding 示例
 

package/lib/drama/skills/short-drama-review/references/production-quality-gates.md

@@ -10,8 +10,8 @@
 - [审查问题](#六审查问题怎么写)
 
 来自对完整生产链返工原因的定性案例阅读，示例均为合成改写。审查方法见
-[review-method.md](review-method.md)。本文展开 `REV-08` 与 `REV-09`：检查具体
-必须保留的内容是否还在，而不是用样本规模、固定数量或所谓命中率证明质量。
+[review-method.md](review-method.md)。按 `REV-08` 与 `REV-09` 核对必须保留的具体内容，
+用来源与产物的差异说明问题；样本规模、固定数量或命中率不能代替内容证据。
 
 **每个环节都会成为下一环节的输入。审查必须指出丢失了哪项故事意义、状态、原文内容
 或制作要求，以及应由哪个技能修复。**
@@ -39,15 +39,15 @@
 检查：
 
 - 是否能从剧本的每个制作相关段落直接追到承担它的镜头或声音；
-- 每个有明确来源的动作是否只实现一次；
+- 每个有明确来源的动作是否只有一个主要落实镜头；其他镜头是否增加反应、细节或重新解释；
+- 若采用 `SHT-08` 允许的峰值冲击重复，是否明确指向主要镜头、各次重复使用同一前后状态，且故事状态只改变一次；
 - 关键结果落地后，观众是否能看到受影响者的反应或后果；
 - 重新编组前后，原文段落、镜头目的、起点/终点与声音引用是否仍相同。
 
 ### 2. 表演只有标签
 
 “震惊、愤怒、电影感”没有触发、处理、选择与落点，删掉台词后只剩静态姿势。
-修复不是堆更多同义词，而是把故事变化落成注视对象、手部选择、重心、停顿、距离、
-呼吸或道具处理。
+修订目标是让故事变化落成注视对象、手部选择、重心、停顿、距离、呼吸或道具处理。
 
 ### 3. 空间与状态漂移
 
@@ -86,10 +86,10 @@ H3 无非画内配乐时，`non_diegetic_music: N/A` 是有效的空形式。
 
 ### 6. 版本和执行状态冒充内容权威
 
-同一集、镜头或提示词可能保留多个历史版本。审查前先读取当前 creator-visible 文档，并按文件名、
+同一集、镜头或提示词可能保留多个历史版本。审查前先读取创作者当前可见的文档，并按文件名、
 标题 ID 和文中明确写出的母版/替代关系判断当前版本；文件更新时间、编号大小、默认展示项或最近一次
 执行状态只能帮助定位，不能替代创作者当前决定。下游仍引用旧标题时，即使内容看似合理也要报告
-版本引用过期，而不是把旧稿重新解释成现行要求。
+版本引用过期，要求负责技能更新引用。
 
 生成或处理任务的技术状态只说明某次执行结束到哪一步，不证明语义质量。任何状态为成功的
 历史记录都必须写明当时准确的提示词标题、资产/镜头引用与配置；审查仍比较这些文字是否
@@ -133,10 +133,10 @@ H3 无非画内配乐时，`non_diegetic_music: N/A` 是有效的空形式。
 不能证明提示词覆盖了原动作、反应、台词与已确认的项目指令。审查者必须比较
 改写前后必须保留的内容。
 
-缩短修改若形成补拍或替代版，必须在 `视频提示词.md` 中标明版本类型、同一文件内的母版或补充
-`MOTION-...` 标题，并把每项原文要求对应到当前版本或说明去向。普通母版不为“没有版本关系”制造
-占位账目。补拍默认不替代母版；替代版只能提出替代请求，最终由审查结论直接指名保留哪一个标题，
-不能让两个版本互相声明对方才是当前版本。若有内容丢失，再由
+缩短修改若形成补拍或替代候选，必须在 `视频提示词.md` 的普通说明中标明版本关系，并把每项
+原文要求对应到当前版本或说明去向。同一 `SHOT-...` 只保留一个正式 `MOTION-...` 及一份可复制正文；
+独立补拍须先由分镜建立对应 `SHOT-...`，再引用其 `MOTION-...` 标题。普通母版不增加版本账目。
+补拍默认不替代母版；替代候选只提出请求，由审查结论明确是否采用，再由负责技能更新正式正文。若有内容丢失，再由
 分镜或视频提示词技能决定保留其他镜头、拆镜、延长、删减或请求上游修改。
 
 ## 四、各环节证据问题
@@ -156,7 +156,8 @@ H3 无非画内配乐时，`non_diegetic_music: N/A` 是有效的空形式。
 - 每项细节是否服务识别、复用、状态、功能或连续性？
 - 身份与造型、地点与观察方向、道具与状态是否分开？
 - 白底/方位/产品板是否写入了只属于具体镜头的情绪、构图或光线？
-- `exact/graphic/no-text` 文字政策是否与提示词一致？
+- 资产的 `exact_readable / graphic_only / no_readable_text / pending_creator_text` 政策，是否与本次提示词的
+  `readable / symbolic / blank / postproduction` 呈现方法一致？前者规定资产事实，后者规定本次图片怎样承载。
 - 输入参考图的可见文字也符合该政策吗？不符合时是否记录了检查和处理方法，而不是指望
   `no-text` 删除参考像素？
 - 每张参考图的 `role`、可参考内容与不可照搬内容是否明确？构图、尺度或效果参考有没有
@@ -164,7 +165,7 @@ H3 无非画内配乐时，`non_diegetic_music: N/A` 是有效的空形式。
 - 像素、文字、水印或裁切检查是否有准确的观察记录；没有时是否明确写“未观察”，
   而不是从文件引用或负面词推断已经看过输入参考图？
 - 定点修改是否写准确目标、`changes` 与 `preserve_set`？
-- 同一地点在同一时间/天气 state 下的各观察方向板，主光来源、冷暖主次、光比方向与实际
+- 同一地点在同一时间/天气状态下的各观察方向板，主光来源、冷暖主次、光比方向与实际
   灯具开关是否对齐（`IMG-10`）？不一致时是否有已记录的成因与对应 delta？逐板单独看
   都成立不能作为通过依据，必须并排比同组板。
 
@@ -201,11 +202,11 @@ H3 无非画内配乐时，`non_diegetic_music: N/A` 是有效的空形式。
 - 缺逐字台词：只阻断具体时长、台词表演与口型可行性，不能自编台词补洞；
 - 只有签名职责、没有正式签名字形：只阻断该承载面的 `readable` 文字投产；若已接受文字政策
   允许 `postproduction`，仅预留承载区可以继续，不能臆造笔迹；
-- 没有创作者声明的播放面遮挡区：只阻断播放面避让核验与整体 `delivery-ready`，画面候选照常
-  预览，不能拿通用安全框代替。
+- 本次交付明确要求核验播放面避让、但缺少创作者声明的遮挡区：只把这项核验记为未完成，
+  不猜通用安全框；普通项目未声明播放面时，`SHT-15` 保持不生效。
 
-这些状态证明的是输入和文本准备度，不证明媒体结果质量；只有全部适用范围都为 `ready`，
-才可声称整体 `delivery-ready`。输入补齐后只重审受影响范围。
+这些判断只说明当前输入与文本是否足够，不证明媒体结果质量，也不要求另建准备度档案。
+本次已启用的必要交付判断缺少输入时，明确列出尚不能确认的范围；输入补齐后只重审受影响部分。
 若镜头写着空白承载区却让人物认出该签名/文字，直接报告语义矛盾；不能用“后期再加”倒推
 当前帧已经具备角色接收和观众阅读所需的证据。
 
@@ -234,8 +235,8 @@ H3 无非画内配乐时，`non_diegetic_music: N/A` 是有效的空形式。
   时间线层"的已接受方案冲突？成品里究竟有没有烧进音轨**不能由本环节判断**——只有已授权的
   文字观察记录报告了该问题时才据此写 finding，否则保持 `unverified`，不从文本推断混音结果。
 - 参考帧已有的外观是否被无谓重复，反而增加漂移？
-- 补拍版或替代版是否把每项原文要求分别对应到运动字段或说明去向，并写明
-  母版或补充运动记录 ID；若请求替代母版，下游审查结论是否完成覆盖与连续性复核？
+- 补拍或替代候选是否把每项原文要求对应到当前版本或说明去向，且正式 `SHOT-...` / `MOTION-...`
+  仍一一对应；独立补拍是否先有相应分镜？请求替代母版时，审查结论是否完成覆盖与连续性复核？
 
 ## 五、生产纪律
 
@@ -248,10 +249,10 @@ H3 无非画内配乐时，`non_diegetic_music: N/A` 是有效的空形式。
 4. **文本策略双重核对**：提示词约束与资产文字政策都要一致；后处理能力不能
    为矛盾提示词开脱。
 5. **定点修改一次改一个意图**：同时显式冻结不变项，记录该修改影响哪些下游。
-6. **重试按被拒的那一项路由**：每次重投先写清这次改了什么。拒绝原因指向某项输入时，
-   先改那一项再投——指向台词或提示词文字就改写该句，指向参考图就换图或修掉图上被拒
-   的部分，指向音频就重录或换音源。被拒的台词改写成同样完成这一拍的另一种说法，
-   不是把这一拍删掉。只有拒绝原因不落在任何具体输入上时，原样重投才算一次合理尝试，
+6. **重试按被拒的那一项路由**：修订要求先写清要改什么。拒绝原因指向台词时交写作负责人，
+   指向提示词文字时交提示词负责人，指向参考图时交参考负责人换图或修图，指向音频时交声音负责人重录或换源。
+   台词的新说法仍须完成同一拍的戏剧功能。审查不代改来源，重投由生产技能展示精确任务预览并取得明确确认。
+   只有拒绝原因不落在任何具体输入上时，原样重投才算一次合理尝试，
    同一输入连续原样重投不超过两次，之后改内容或退回负责环节。审查问题写在该改的那项
    输入上。每次记录失败、改动与 `preserve_set`。
 7. **长范围先建连续性记录和检查点**：可一次做完整原文覆盖，但接受、审查与回滚仍按

package/lib/drama/skills/short-drama-review/references/project-calibration.md

@@ -1,6 +1,6 @@
 # 项目级试跑与生产反馈校准
 
-校准只消费创作者提供或授权形成的**文字观察记录**，用来改进当前项目的提示词与镜头设计。
+校准只读取创作者提供或授权形成的**文字观察记录**，用来改进当前项目的提示词与镜头设计。
 本技能不提交任务、不下载或生成媒体，也不从“任务成功”、结果地址或文件存在推断内容质量。
 
 ## 先分清两种观察
@@ -11,24 +11,24 @@
 
 两种记录都要写明被观察的媒体、观察方法、准确的文档标题引用、观察区域/时间段、限制与观察者。
 生成结果还要写明对应提示词或镜头标题、参考图顺序、制作形态/配置和结论适用范围。审查只读取
-项目提供的记录，不在公开运行时找数据。记录仍可使用只代表这份有界观察可复核，不表示创作内容
-已接受或质量通过；证据被撤回、媒体已经换过或来源无法复核时，不沿用它做校准，并写明失效原因。
+项目提供的记录，不自行寻找其他项目数据。记录可复核，不代表创作内容已接受或质量通过；
+证据被撤回、媒体已经换过或来源无法复核时，停止引用该记录并说明原因。
 
-实际项目记录可以由创作者放在受保护的 `原始资料/生产观察.md`，也可以直接在当前请求中提供；它不属于
+实际项目记录可以由创作者放在受保护的 `输入/生产观察.md`，也可以直接在当前请求中提供；它不属于
 五份单集创作文档，也不进入文本交付包。
 
 ## 选择有内容代表性的试跑
 
 不随机抽样，也不规定固定数量。按项目实际风险选互相有差异的代表：身份近景与全身、双人潜台词、
 多人同框、手部/道具接触、可读文字、大动作/复杂路径、高难度光影、首尾帧接续，以及最重要的
-lookdev 风格帧。选择理由必须对应本项目，不因别处高频而加入。
+lookdev 风格帧。选择理由必须对应本项目的风险。
 
 ## 从观察到诊断
 
 逐条区分：
 
 1. **直接观察**：被授权观察者实际看见或听见什么；
-2. **文本定位**：哪个 prompt/spec/shot 字段可能参与；
+2. **文本定位**：哪个提示词、规格或镜头条目可能参与；
 3. **替代解释**：参考本身、景别、动作竞争、配置或观察盲区是否也可能造成结果；
 4. **有界结论**：只在当前项目与准确版本条件下，哪种写法更可读或更稳定；
 5. **修订路由**：身份/地理回 assets，构图/镜头边界回 storyboard，图片规格回 image-prompts，
@@ -38,37 +38,38 @@ lookdev 风格帧。选择理由必须对应本项目，不因别处高频而加
 
 ## 诊断之后先定处置，再谈修订
 
-拿到诊断不等于要重做。先在五种处置里选一种，再决定要不要进入修订；顺序反了就会把最贵的
-一种当成默认：
+诊断之后先选择处置，再决定是否修订：
 
 | 处置 | 什么时候选 | 本环节要做什么 |
 |---|---|---|
 | 保留 | 镜头目的已达成，剩余瑕疵不影响信息、连续性与观众理解 | 不改文字；把已知瑕疵记进观察，不伪装成通过 |
 | 后期处理 | 缺陷落在文字层、混音、调色、修剪或少量边缘帧上 | 不改提示词；交接给对应责任方并记录归属 |
 | 局部编辑 | 构图、时机与表演都对，只有一层错 | 只写该层的修改项与保持项，不重写整条提示词 |
-| 重新提交 | 文字经核对没有缺陷，失败看起来随机 | 有上限的尝试，不是修复手段 |
+| 重新提交 | 文字经核对没有缺陷，现有证据尚未指向可控原因 | 交负责生产的技能提出有上限的尝试，保留本次任务预览与明确确认 |
 | 改写 | 同一缺陷在多次结果中重复出现 | 回到诊断指出的那一层改结构：拆镜、减并发、换实现路径 |
 
 两条判据把它们分开：
 
 1. **缺陷是否落在文字能控制的范围内**。 落在文字之外的（渲染瑕疵、编码、少数边缘帧）不该
    由提示词承担；改提示词只会引入新变量，还会让下一次结果更难归因。
-2. **同一缺陷是否已经复现过**。 第一次出现可能是随机，重复出现说明结构有问题。此时继续
-   重新提交，是在为同一个错误反复付费，而每次都不产生新信息。
+2. **同一缺陷是否已经复现过**。重复出现时重新核对参考、文本与配置等解释；诊断指向结构后，
+   应交负责技能修改该结构，不能继续把同样的重投当作修复。
 
 **不要用增加形容词的方式回应重复缺陷**。 在同一条提示词上叠加“高质量、避免变形、保持一致”
 一类词，既没有改变任何可执行事实，又让下一次失败失去可比性：文本变了，但没人说得清变的是
 哪一条约束。若诊断指向结构，就改结构。
 
-处置要写进观察或审查记录，说明选了哪一种、依据是上面哪一条。不写处置直接给修订，等于默认
-所有缺陷都该由提示词修——这通常是最贵的假设，也是把“抽卡”当成方法的起点。
+处置写进观察或审查记录，说明选择依据。保留或交后期处理时，不虚构提示词改动；
+选择局部编辑、重投或改写时，写清负责方须完成的动作。
 
-只有选择了局部编辑或改写、且缩小了原文要求范围时，才建立补拍或替代版；它们的记账方式见
-[阶段契约](stage-contract.md) 的参考媒体与补拍一节。
+选择局部编辑或改写、且缩小了原文要求范围时，修订请求须交代补拍或替代关系：在当前
+`视频提示词.md` 的普通说明中引用母版，说明每项原文要求由当前版本、母版、其他补拍承担，还是退回
+分镜修订。同一 `SHOT-...` 仍只对应一个正式 `MOTION-...` 和一份可复制正文；独立补拍先交分镜建立
+相应镜头，再保留与母版的关系。替代候选只提出请求，由后续审查明确结论，再交负责技能更新正式正文。
 
 ## 单变量有界修订
 
-每次只改变当前诊断需要的变量；必须联动的变量一起说明。修订 finding 至少包含准确目标 refs、
+每次只改变当前诊断需要的变量；必须联动的变量一起说明。修订 finding 至少包含准确目标引用、
 诊断、`change_set`、`preserve_set`、联动理由、预期可观察变化和仍未知内容。保留剧本原文覆盖、角色
 身份、镜头目的、信息权限和终点义务。同一失败不能靠重复提交完全相同的提示词解决。
 

package/lib/drama/skills/short-drama-review/references/review-method.md

@@ -40,7 +40,7 @@ Mechanical examples:
 Semantic examples:
 
 - downstream action changes story meaning;
-- a scene has no opposition or turn;
+- a conflict scene has no opposition or turn, or another kind of scene lacks a necessary state change;
 - a keyframe implies several moments;
 - action load crowds out performance;
 - escalation repeats the same mechanism;
@@ -54,7 +54,7 @@ A useful finding contains:
 2. project-relative file plus heading ID, line number or shot when useful;
 3. bounded quotation or conflicting facts;
 4. audience, continuity or production impact;
-5. required outcome and preserve set, not ghostwritten replacement prose;
+5. required outcome and the facts that must stay unchanged, leaving replacement prose to the owner;
 6. owner, severity and open/closed status.
 
 “The dialogue is weak” is invalid. A finding should identify the exact exchange, say what fails to change, and tell
@@ -73,11 +73,13 @@ Resolve upstream meaning first; do not polish prompt wording when the shot or as
 Group duplicate findings and route them to the owning skill. A revision request states the target outcome, preserve
 set and review scope to rerun.
 
-`REV-14`: a revision request is a patch, not a rewrite. It names the smallest change at a cited location; local
-findings never add up to a request to rewrite the episode. When the review compares a revision with the version before
+`REV-14`: a revision request names the smallest change at a cited location and the facts to preserve. Local
+findings do not authorize a whole-episode rewrite. When the review compares a revision with the version before
 it, any strong hook, payoff line, or comic beat that polishing removed is listed under `## 被删的强钩子` with its prior
-location and a short quote, so the owner can restore it. The reviewer lists it; the owner decides. On a later explicit re-review request, read the current version, confirm the finding is addressed, check the
-preserve set and close or retain each affected finding. A different reviewer is useful but not mechanically required.
+location and a short quote, so the owner can decide whether to restore it.
+
+On a later explicit re-review request, read the current version, confirm the required outcome, check the preserve
+set and close or retain each affected finding. Prefer another uninvolved reviewer when available; otherwise disclose self-review.
 
 ## Anti-template review