- name
- techvideo-wizard-skill
- description
- Interactive video script wizard that guides users through a structured 6-step process to collect product video requirements and generate a professional video script in Markdown format. This skill should be used when users want to write a video script, need guidance on creating product demo or promotional videos, express intent like "help me write a XXX video script", "video script wizard", or when any task requires structured requirement gathering for video/multimedia content creation. Typical trigger examples include "帮我写个 SuperMap GIS 小视频的脚本", "帮我写个产品介绍视频的脚本", "视频脚本向导".
TechVideo Wizard - Video Script Generator
Overview
Guide users through a structured 6-step interactive process to collect product/feature video requirements, then automatically generate a professional video script. The output is a Markdown-formatted script file optimized for UI/design team collaboration.
Core value: Transform vague ideas into production-ready video scripts through progressive refinement.
Workflow Protocol
Step 1: 一句话介绍 (必填) → [自动] 重复内容检查 → Step 2: 历史 vs 当前对比 (可选) → Step 3: 核心步骤 (必填) → Step 4: 变化/效果 (可选) → Step 5: 最终效果 (必填) → Step 6: 生成脚本 (自动) → [可选] Step 7: 自动优化Interaction rules:
- Ask exactly one question per step using
ask_followup_question - Wait for user response before proceeding to the next step
- Store each answer as context for subsequent steps
- Step 2 and Step 4 are optional — accept skip keywords ("无", "跳过", "没有", "skip", "none") and proceed directly to the next step
Step 1: One-Line Introduction (Guided + Strong Constraints)
Collect the user's core intent in a single, visualizable sentence.
Use the following prompt verbatim:
你是一个视频脚本助手,需要引导用户填写视频内容。 请用户输入"一句话介绍"。 【填写要求】 1. 使用"动词 + 对象 + 结果特征" 2. 必须具体、可视化(能让人脑海有画面) 3. 避免抽象表达(如:提升效率、优化体验) 4. 每句话不超过 15 字 【示例】 - 构建高真实感铁路三维场景 - 实现网络故障自动定位 - 自动识别图像中的异常目标 - 快速生成标准化运维报告 - 实现跨系统统一调度 【错误示例】 - 提升效率 ❌ - 优化用户体验 ❌ 请直接给出一句话介绍:
Processing the answer:
- Validate that the response follows "动词 + 对象 + 结果特征" pattern
- If too abstract (contains words like 提升、优化、增强 without specifics), prompt the user to refine with concrete imagery
- Verify each sentence ≤ 15 characters
- Save as
INTRODUCTIONcontext variable - IMMEDIATELY proceed to Duplicate Content Check (see below)
Duplicate Content Check (Auto-trigger after Step 1)
After collecting the one-line introduction, automatically check if similar video content already exists on WeChat Video Channels (微信视频号), Bilibili (B站), and Douyin (抖音).
Execution logic:
- Extract search keywords from
INTRODUCTION - Use
web_searchtool to search for existing content (execute multiple searches in parallel):
Search queries: - {关键词} 微信视频号 - 微信视频号搜索 - {关键词} 视频号 - 视频号通用搜索 - {关键词} site:bilibili.com - B站官方域名搜索 - {关键词} B站 视频 - B站通用搜索 - {关键词} site:douyin.com - 抖音官方域名搜索 - {关键词} 抖音 视频 - 抖音通用搜索
- Present search results to user with clear formatting:
🔍 重复内容检查完成 微信视频号搜索结果: - ✅ 视频标题 - 平台:微信视频号 | 发布时间:YYYY-MM-DD - (无结果) B站搜索结果: - ✅ 视频标题 - 平台:B站 | 时长:XX:XX | 发布时间:YYYY-MM-DD - (无结果) 抖音搜索结果: - ✅ 视频标题 - 平台:抖音 | 发布时间:YYYY-MM-DD - (无结果) 补充检查:如需进一步确认,请在以下文档中使用关键词搜索是否存在相似视频: 视频列表文档 是否继续制作? - 回复「继续」→ 进入 Step 2 - 回复「调整主题」→ 重新思考视频角度
Processing rules:
- If user chooses "调整主题", restart from Step 1
- If user chooses "继续", proceed to Step 2
Search optimization tips:
- Use specific product/feature names for better results
- If no results found on first try, use alternative keywords
- Prioritize recent content (within 1 year) for relevance
Step 2: Before vs After Comparison (Optional)
Collect optional historical vs current state comparison to highlight the transformation value.
Use the following prompt verbatim:
请描述:有了这个功能后,效果有哪些变化? 【填写要求】 1. 用"原来... → 现在..."的对比方式 2. 量化或具体描述变化 3. 突出价值感 【示例】 - 原来人工建模需要 2 天 → 现在 10 分钟完成 - 原来客户看不到三维效果 → 现在实时在线浏览 - 原来报告生成依赖手工汇总 → 现在一键自动导出 请填写(可选):
Processing the answer:
- Accept skip keywords: "无", "跳过", "没有", "skip", "none", "—", empty response
- If skipped, set
BEFORE_AFTERto empty and proceed to Step 3 - Otherwise save the response as
BEFORE_AFTERcontext variable
Step 3: Core Steps (Key Section)
Break down the requirement into 3-8 actionable steps. This section is the core content of the video.
Use the following prompt verbatim:
【填写要求】 1. 支持 3-8 步(推荐 4-6 步) 2. 每一步包含核心说明和旁白补充(可选) 【格式】 步骤编号:核心说明(≤15字,动词开头) 旁白:讲解内容第一句(≤15字) 旁白:讲解内容第二句(≤15字) (可选)截图:[截图描述或文件名] 【示例】 第一步:导入三维数据 旁白:支持多种格式一键导入 旁白:智能识别坐标自动配准 截图:[数据导入界面] 第二步:配置发布参数 旁白:根据规模智能推荐配置 旁白:拖拽调整实时预览 截图:[参数配置界面] 第三步:一键发布服务 第四步:前端参数调整 【说明】 - 如果只填写"核心说明",AI 会自动生成旁白补充 - 如果填写完整内容,AI 会保持原样 【错误示例】 - 提升分析能力 ❌ - 更高效处理 ❌ 【易懂性审核示例】 输入:"处理数据" 审核:⚠️ 过于笼统,无法形成画面 建议:"读取 Excel 表格数据" 输入:"完成任务" 审核:⚠️ 缺少宾语,什么任务? 建议:"生成数据汇总报告" 输入:"查看结果" 审核:⚠️ 歧义:是图表、报告还是列表? 建议:"查看预测准确率图表" 请按"每行一步"填写:
Processing the answer:
- Parse steps with structure:
步骤编号,旁白:,截图: - Step-by-step review: After user submits each step, apply the "User Comprehension Review"
User Comprehension Review (Each Step)
After user inputs each step, evaluate from the target audience's perspective:
Review criteria:
- 具体性: 是否具体可感知?还是有歧义?
- 画面感: 用户脑海里能否形成画面?
- 易懂性: 是否有过多专业术语?普通用户能理解吗?
- 动词明确性: 动词是否清晰表达动作?
If issues found, show:
📝 文字优化建议 【原文】xxx 【问题】xxx(如:专业术语过多、不够具体、可能有歧义) 【建议修改】xxx 是否接受修改? - 回复「接受」→ 使用优化后的文字 - 回复「保持」→ 保留原样 - 回复「其他建议」→ 提供你自己的想法
Narration Design Principle (CRITICAL):
When generating or optimizing narration, follow this logic pattern:
- 先确定目标 - 要操作什么、在哪里操作
- 再指定来源 - 数据从哪来、选什么
- 最后执行 - 确认后执行操作
Example:
- User input: "导入符号库"
- Narration: 先打开符号库管理界面 → 再导出待合并工作空间中的符号库 → 最后统一导入目标工作空间
Apply this logic to ALL steps.
If text is clear, acknowledge and continue:
✅ 收到,继续下一步(或请填写下一步)
- If step has no narration, generate 2-3 sentences based on core description (after user confirms)
- Each generated narration ≤ 15 characters
- Save as
CORE_STEPScontext variable (preserve structure for script generation)
Step Review & Modification (After All Steps Collected)
After all steps are collected (用户完成所有步骤填写后), display a summary and ask for modification preference:
✅ 所有步骤已收集完成! 当前步骤概览: 1. 第一步:导入三维数据 2. 第二步:配置发布参数 3. 第三步:一键发布服务 ... 【过渡提示 - 随机选择以下示例之一】 - "好了,以上就是XX的配置方法" - "这样,XX就安装完成了" - "按照这三步,XX就能正常跑起来了" - "完成这些操作,XX就设置好了" - "通过以上步骤,XX已经准备就绪" 是否需要修改? - 回复「修改第X步」→ 重新编辑指定步骤(如:修改第2步) - 回复「修改全部」→ 重新编辑所有步骤 - 回复「继续」→ 进入下一步骤,描述最终效果
Modification handling:
- 修改第X步: Show the selected step content and ask for new input, then return to summary
- 修改全部: Clear
CORE_STEPSand restart Step 3 from beginning - 继续: Proceed to Step 4
Step 4: Change/Effect (Optional Guidance)
Collect optional before/after comparison information.
Use the following prompt verbatim:
请描述这个功能带来的变化或效果(可选)。 【建议】 用"原来 vs 现在"的方式表达,更容易理解 【示例】 - 原来人工逐点排查 → 现在系统自动定位 - 原来耗时数小时 → 现在分钟级完成 - 无需人工建模 → 自动生成结果 【说明】 如果这个功能本身已经很直观,可以简单描述或跳过 请填写(可选):
Processing the answer:
- Accept skip keywords: "无", "跳过", "没有", "skip", "none", "—", empty response
- If skipped, set
CHANGE_EFFECTto empty and proceed to Step 5 - Otherwise save the response as
CHANGE_EFFECTcontext variable
Step 5: Final Effect (Strong Constraint)
Collect quantifiable, perceivable outcome metrics.
Use the following prompt verbatim:
请描述最终带来的效果。 【填写要求】 1. 必须具体、可感知 2. 尽量量化或对比 3. 禁止使用空话(如提升效率、优化体验) 【示例】 - 故障定位时间缩短80% - 建模时间从2小时降至10分钟 - 操作步骤由5步减少至1步 - 人工成本降低50% 【错误示例】 - 提升效率 ❌ - 优化体验 ❌ 请填写:
Processing the answer:
- Enforce specificity — reject abstract responses
- Look for numbers, percentages, time comparisons, or concrete metrics
- Save as
FINAL_EFFECTcontext variable
Step 6: Generate Video Script (Auto Output)
Automatically generate the final video script based on all collected context.
Do NOT ask the user anything in this step. Execute the following generation logic:
Terminology Constraints (CRITICAL)
Before generating, read and apply constraints from TERM_CONSTRAINTS.md:
- Load constraints: Read the terminology constraint table
- Preserve terms: All terms listed in the constraint table must be preserved exactly as written
- Priority rule: Terminology constraints override general optimization rules
- User-defined terms: If user explicitly states "请保留XX术语", add to temporary constraint list
Constraint categories:
- Product Names (e.g., SuperMap GIS, SuperMap iDesktop)
- Technical Terms (e.g., 三维场景, 倾斜摄影, BIM模型)
- Feature Modules (e.g., 数据导入, 服务发布)
- Industry Applications (e.g., 智慧轨交, 数字孪生)
- Action Commands (e.g., 一键导入, 实时预览)
- Effect Descriptions (e.g., 纤毫毕现, 身临其境)
Generation Template
Use this structure to generate the script:
# {基于 INTRODUCTION 的视频标题} - 视频脚本
> **生成时间**: {当前日期}
> **风格**: 科技感、简洁、有节奏
---
## 脚本正文
### 🎬 开头
{基于 INTRODUCTION 生成 1-2 句开场文案,每句 ≤ 15 字}
{**要求**:必须包含问句或痛点场景,吸引注意力}
{**示例**:"还在为数据导入报错发愁吗?" 而非 "这是 GIS 数据导入功能"}
### 🔧 方案
{基于 CORE_STEPS 生成详细分镜文案,每个步骤包含:}
{1. 步骤标题(≤15字)}
{2. 旁白内容(2-3句,每句 ≤ 15 字)}
{3. 截图位置标注(如果有)}
{如果用户填写了旁白,保持原样;如果未填写,AI 根据核心说明拓展生成}
{如果 BEFORE_AFTER 非空,插入对比说明:}
**前后对比(可选)**
{基于 BEFORE_AFTER 生成 1-2 句对比文案}
### 📈 结果
{基于 FINAL_EFFECT 生成 1-2 句结果文案,每句 ≤ 15 字,量化价值}
{**过渡提示**:随机选择以下示例之一}
{"好了,以上就是XX的配置方法" / "这样,XX就安装完成了" / "按照这三步,XX就能正常跑起来了" / "完成这些操作,XX就设置好了" / "通过以上步骤,XX已经准备就绪"}
### 🎯 收尾(包含行动号召 CTA)
{1-2 句收尾文案,每句 ≤ 15 字}
{**要求**:必须包含明确的行动号召 (CTA),引导用户互动}
{**CTA 示例**:}
{- "想亲自试试?评论区领安装包。"}
{- "觉得好用,点个赞支持一下。"}
{- "你的 GIS 项目还在手工建模吗?"}
{- "遇到问题?评论区告诉我。"}
{- "想看更多技巧?关注不迷路。"}
---
## 原始输入记录
| 项目 | 内容 |
|------|------|
| 一句话介绍 | {INTRODUCTION} |
| 历史 vs 当前对比 | {BEFORE_AFTER 或 "(已跳过)"} |
| 核心步骤 | {CORE_STEPS} |
| 变化/效果 | {CHANGE_EFFECT 或 "(已跳过)"} |
| 最终效果 | {FINAL_EFFECT} |Output Rules
- Structure: 开头(含问句/痛点) → 方案 → 结果(含过渡提示) → 收尾(含 CTA)
- Line length: Each sentence MUST be ≤ 15 Chinese characters
- Narration generation: If user provided narration, keep it; otherwise generate 2-3 sentences per step
- Style: Tech-forward, concise, rhythmic — prioritize "value perception"
- No repetition: Avoid repeating the same concept across sections
- Value-first: Every line should convey tangible benefit or capability
- Terminology preservation: Strictly follow
TERM_CONSTRAINTS.md— preserve all fixed terms exactly as specified - Opening requirement: 开头 MUST contain a question or pain point scenario (e.g., "还在为数据导入报错发愁吗?")
- CTA requirement: 收尾 MUST include clear Call-to-Action to guide user interaction (e.g., "想亲自试试?评论区领安装包。")
File Output
After generating the script content:
- Write the complete Markdown content to
{视频标题}-视频脚本.mdin the current working directory usingwrite_to_file - Display a summary showing: file path, total lines
- Ask if user wants Step 7 optimization
Step 7: Auto-Optimize Input (Optional Enhancement)
Offer to optimize/refine the raw inputs for better video expression.
Only execute if user agrees after Step 6.
Use the following prompt:
是否需要对原始输入内容进行自动优化? 优化将使表达更清晰、更具体、更适合视频旁白表达。
If user confirms, apply this transformation to each of the 4 input fields:
Optimization template:
请将以下内容优化为更具体、可用于视频表达的句子: 要求: 1. 更清晰 2. 更具体 3. 避免抽象词(如提升、优化) 【输入】 {{用户填写的原始内容}}
Apply optimization to INTRODUCTION, BEFORE_AFTER, CORE_STEPS, CHANGE_EFFECT, and FINAL_EFFECT respectively, then regenerate the script with optimized content. Save as {视频标题}-视频脚本-优化版.md.
Optimization Constraints:
- All terms from
TERM_CONSTRAINTS.mdmust be preserved exactly - Optimization should improve clarity and flow without changing fixed terminology
- When terminology conflicts with length constraints, preserve the term and adjust surrounding text
Anti-Patterns
DO NOT:
- Ask multiple questions in a single step
- Skip Step 1, Step 3, or Step 5 (these are mandatory)
- Use sentences exceeding 15 characters in the output script
- Include abstract terms like "提升效率", "优化体验" in the generated script
- Proceed to Step 6 without collecting answers from all mandatory steps
DO:
- Validate each answer against the constraints before proceeding
- Gently guide users to rephrase if their input violates the format rules
- Maintain a helpful, guiding tone throughout
- Preserve exact prompt wording from Steps 1-5 as specified above
- Automatically trigger duplicate content check after Step 1
Quality Checklist
Before delivering the final script, verify:
- [ ] All 3 mandatory fields collected (Step 1: INTRODUCTION, Step 3: CORE_STEPS, Step 5: FINAL_EFFECT; Step 2 and Step 4 may be skipped)
- [ ] Each output sentence ≤ 15 characters
- [ ] Script follows 开头→方案→结果→收尾 structure
- [ ] Each step has core description + narration (user-provided or AI-generated)
- [ ] No abstract/empty phrases in output
- [ ] All terms from
TERM_CONSTRAINTS.mdpreserved exactly - [ ] Markdown file written to working directory
- [ ] File name follows
{title}-视频脚本.mdformat - [ ] Duplicate content check performed after Step 1