Back to Details

reference-decode

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Reference Decode

参考解码

Follow shared release-shell rules in:
  • postplus-shared
    release-shell rules
Use this skill before prompt writing when any source material or rough concept exists.
This skill is for:
  • breaking a good reference into reusable logic
  • separating what should be copied from what must not be copied
  • turning vague ideas into promptable structure
  • creating a stable handoff for storyboard or provider request work
This skill is not for making the final storyboard grid or provider payload.
遵循
postplus-shared
发布框架中的共享规则：
  • postplus-shared
    发布框架规则
当存在任何源素材或粗略概念时，在撰写提示前使用此技能。
此技能适用于：
  • 将优质参考拆解为可复用逻辑
  • 区分应当复刻与绝对不能复刻的内容
  • 将模糊想法转化为可生成提示的结构化内容
  • 为分镜制作或供应商需求工作创建稳定的交接流程
此技能不适用于制作最终分镜网格或供应商交付内容。

Core Rule

核心规则

Do not summarize references as "good vibe", "nice pacing", or "strong chemistry".
Extract four objects:
  1. hookEssence
  2. viewerQuestion
  3. mustCopyVisualGrammar
  4. forbiddenDrift
不要将参考总结为“良好氛围”“节奏不错”或“化学反应强烈”这类表述。
需提取四个要素：
  1. hookEssence
  2. viewerQuestion
  3. mustCopyVisualGrammar
  4. forbiddenDrift

Default Workflow

默认工作流程

1. Load the smallest sufficient reference set

1. 加载最小且足够的参考集

Prefer:
  1. hook-first clip or first 0-5 seconds
  2. hook-first contact sheet
  3. supporting benchmark note
  4. full-video style board only if hook-first material is insufficient
If the task is specifically hook replication, do not default to downstream style boards or product-demo frames.
优先选择：
  1. 钩子前置片段或开头0-5秒内容
  2. 钩子前置联系表
  3. 辅助基准说明
  4. 仅当钩子前置素材不足时，才使用全视频风格板
如果任务是专门的钩子复刻，不要默认使用下游风格板或产品演示帧。

2. Decode the opening mechanism

2. 解码开场机制

Ask:
  • what is the first clear promise?
  • what question enters the viewer's head?
  • what exact visual structure makes that promise legible?
思考：
  • 第一个明确的承诺是什么？
  • 观众脑海中会产生什么疑问？
  • 是什么具体的视觉结构让这个承诺清晰可辨？

3. Separate structure from identity

3. 区分结构与标识

Keep:
  • camera grammar
  • shot order
  • visible object logic
  • timing logic
  • relationship logic
Do not keep:
  • exact faces
  • exact wardrobe
  • creator identity
  • exact text overlay
  • exact location
保留：
  • 镜头语法
  • 镜头顺序
  • 可见物体逻辑
  • 时序逻辑
  • 关系逻辑
不要保留：
  • 具体人脸
  • 具体服装
  • 创作者身份
  • 具体文字叠加层
  • 具体地点

4. Print the decode block before writing prompts

4. 撰写提示前先输出解码模块

Read 
references/output-shape.md
.
Print the decode in that exact shape first.
阅读
references/output-shape.md
。
先严格按照该格式输出解码内容。

5. Hand off cleanly

5. 清晰交接

  • if the target is a grid, hand off to 
    storyboard-grid-writer
  • if the target is a full render request, hand off to 
    video-request-architect
  • if references should be explicitly bounded, hand off to 
    reference-contract-builder
  • 如果目标是分镜网格，交接给
    storyboard-grid-writer
  • 如果目标是完整渲染请求，交接给
    video-request-architect
  • 如果需要明确限定参考范围，交接给
    reference-contract-builder

No-Reference Rule

无参考规则

If there is no usable reference, you may still use this skill in proxy mode.
In proxy mode:
  • start from the chosen segment pattern
  • infer the likely viewer question
  • write must-copy grammar as required scene anchors, not as imitation notes
  • write forbidden drift as anti-generic safeguards
Make it explicit that the output is derived from the brief, not from observed benchmark footage.
如果没有可用参考，你仍可以在代理模式下使用此技能。
在代理模式中：
  • 从选定的片段模式入手
  • 推断观众可能产生的疑问
  • 将必须复刻的语法写为所需场景锚点，而非模仿说明
  • 将禁止偏离的内容写为反通用防护措施
需明确说明输出内容源自任务简报，而非观察到的基准素材。

Failure Mode

失败模式

Stop and say the decode is under-specified if there is no way to determine:
  • what the opening promise is
  • what the viewer should notice first
  • what the scene must visibly contain
Do not move to storyboard writing with only mood words.
如果无法确定以下内容，请停止并说明解码规格不足：
  • 开场承诺是什么
  • 观众首先应注意到什么
  • 场景必须包含哪些可见元素
不要仅凭借情绪词汇就开始撰写分镜。