雀影

AI 视频创意伙伴，生成分镜脚本和 Seedance 提示词

Security Scan

This is an instruction-only creative prompt-engineering skill for generating video storyboards and Seedance prompts; its declared footprint (no installs, no credentials) matches its purpose.

Purpose & Capability

Name/description (video creative partner → generate storyboards and Seedance prompts) align with the skill contents. The files and SKILL.md focus entirely on prompt templates, style guides, and storyboard generation. There are no unrelated env vars, binaries, or config paths requested.

Instruction Scope

SKILL.md asks the agent to read the included规范 documents and to interactively collect/confirm user inputs and assets (images/videos) to produce prompts and deliverables — which is appropriate for this purpose. The allowed-tools list includes Read/Write/Edit so the agent can read bundled docs and user-uploaded assets and write outputs; these actions are expected for a content-generation skill but give the agent the capability to access workspace files. There are no instructions to contact external endpoints, exfiltrate secrets, or read system credentials.

Install Mechanism

Instruction-only skill with no install spec and no code files. No downloads or archive extraction. Lowest install risk.

Credentials

The skill requests no environment variables, credentials, or config paths. All required resources are internal docs and user-supplied media, which is proportionate to the stated functionality.

Persistence & Privilege

always:false and no special privileges. Autonomous invocation is allowed by default. The allowed-tools include Write/Edit which is reasonable for generating/writing the storyboard/prompt outputs, but users should confirm where outputs will be written (dedicated folder) to avoid unintended edits to unrelated files.

Assessment

This skill appears coherent and matches its stated purpose. It does not request credentials or install code. Before installing or using it: 1) Confirm where the skill will write files (set a dedicated output folder). 2) Be mindful about uploading any private or copyrighted assets (the docs reference platform limits and restrictions—e.g., avoid uploading real-person faces if you don't want them used). 3) Review any generated prompts for copyrighted or sensitive content before using them with external services. 4) If you want extra safety, restrict the skill's file access or run it in an environment where it can only read a designated directory. If you want me to, I can summarize what the skill will read/write and produce a short checklist you can use when granting it access.

Domain keywords: 雀影, 视频生成, AI视频, 分镜, Seedance, 即梦, prompt

Summary: 交互式引导完善视频创意，生成分镜脚本和 Seedance 提示词。

Strategy:

1. 用户描述视频内容 → 确认素材、时长、类型
2. 分析故事 → 推荐视觉风格方案
3. 设计角色锚点、镜头节奏、台词音效
4. 确认画面比例 → 生成分镜脚本 → 输出 prompt

AVOID:

• AVOID 不读规范文档就生成，应该先读人物刻画规范、Prompt技术规范
• AVOID 镜头数量超出范围，5s=3-4镜，10s=5-7镜，15s=7-10镜，30s=12-18镜
• AVOID 台词超过6字，每句2-6字，口语化
• AVOID 忘记尾帧处理，最后0.3秒快速切黑

---

规范文档

操作	文档
角色设计	人物刻画规范
Prompt 编写	Prompt技术规范
镜头设计	导演思维与节奏规范
台词音效	台词与音效规范
场景生成	场景生成规范
配音制作	配音制作规范

镜头数量

时长	镜头数
5s	3-4
10s	5-7
15s	7-10
30s	12-18

台词规范

每句 2-6 字，爽感语速每秒 3-5 字，口语化个性化。

输出格式

（时长 X 秒）【标题】（风格）（色调）

【剧作要求】...
【剧情梗概】...
【人物重点】...
【场景重点】...

[镜头 1-首帧] X 秒 景别+运镜 人物描述 动作（场景）
【角色】:"台词" #Xs
（音效）音效 #Xs
...
[镜头 N-尾帧] X 秒 ... 最后 0.3 秒快速切黑