Material Report
Quick Start
- - Request or confirm: video path, category/vertical, target platform.
- Duration spec: keep the new material duration roughly consistent with the original video unless the user explicitly requests otherwise.
- Analyze the provided video and output a markdown report.
Workflow
- 1. Verify the video file exists and read basic metadata (duration, resolution, fps).
- Before extracting frames, check whether frame images already exist. If they do, reuse them directly.
- If frame extraction is needed, use ffmpeg to extract frames. Save all frame images in a folder named after the video (same base name) under the current working directory. If permissions prevent this, ask the user to grant access; do not change the output path.
- If ffmpeg is not available on the user's machine, prompt the user to install ffmpeg (do not provide installation instructions in this skill) or provide an existing frame image folder path.
- Extract the original framework as ordered stages (opening hook, problem, turning points, resolution, ending/CTA).
- Hook identification: prioritize the first 0-5 seconds and any large on-screen headline, abrupt visual change, or explicit "attention" phrasing. If the hook is ambiguous, state multiple candidates and pick the strongest with a brief rationale.
- 6. Summarize material traits (format, tempo, overlays, UI density, meme usage, etc.).
- List acquisition keywords (concise, actionable, aligned to content and visual tactics).
- Create at least one new material framework. For the "different direction" variant, choose the best-performing direction based on your judgment, without being constrained by the original material. Do not deviate from the same category/vertical and target platform.
- Provide a detailed storyboard table for the new framework with time ranges, on-screen action, copy, and visual effects.
- Deliver the report in markdown in the response.
- If the user does not specify category/vertical, first ask for it; then proceed with a best-effort inference and label it clearly as an assumption in the report.
Output Template (Markdown)
- - Load references/report_template.md and follow its structure
Notes
- - Match the user's language and tone.
- Keep the storyboard single-level lists and clear time ranges.
- If the user requests a same-framework version, keep the structure identical and only change content details.
素材报告
快速开始
- - 询问或确认：视频路径、类别/垂直领域、目标平台。
- 时长规范：除非用户明确要求，否则保持新素材时长与原视频大致一致。
- 分析提供的视频并输出一份Markdown报告。
工作流程
- 1. 验证视频文件是否存在，并读取基本元数据（时长、分辨率、帧率）。
- 在提取帧之前，检查帧图像是否已存在。如果存在，则直接复用。
- 如果需要提取帧，使用ffmpeg进行提取。将所有帧图像保存在当前工作目录下以视频名称（相同基础名称）命名的文件夹中。如果权限不允许，请用户授予访问权限；不要更改输出路径。
- 如果用户机器上没有ffmpeg，提示用户安装ffmpeg（本技能不提供安装说明），或提供现有的帧图像文件夹路径。
- 将原始框架提取为有序阶段（开场钩子、问题、转折点、解决方案、结尾/行动号召）。
- 钩子识别：优先关注前0-5秒以及任何大型屏幕标题、突然的视觉变化或明确的注意措辞。如果钩子不明确，列出多个候选方案，并选择最强的一个，附上简要理由。
- 6. 总结素材特征（格式、节奏、叠加元素、UI密度、表情包使用等）。
- 列出获取关键词（简洁、可操作、与内容和视觉策略一致）。
- 创建至少一个新的素材框架。对于不同方向的变体，根据你的判断选择表现最佳的方向，不受原素材限制。但不要偏离相同的类别/垂直领域和目标平台。
- 为新框架提供详细的故事板表格，包含时间范围、屏幕动作、文案和视觉效果。
- 在回复中以Markdown格式交付报告。
- 如果用户未指定类别/垂直领域，先询问；然后尽力推断，并在报告中明确标注为假设。
输出模板（Markdown）
- - 加载参考文件/report_template.md并遵循其结构
注意事项
- - 与用户的语言和语气保持一致。
- 保持故事板为单级列表，并明确时间范围。
- 如果用户要求相同框架的版本，保持结构一致，仅更改内容细节。
