| name | ppt-outline |
| description | Generate a PPT slide outline in Markdown (xxx-outline.md) from user-provided sources such as this repo, local files, pasted context, or fetched web pages. Use when asked to create a slide deck outline / PPT structure / "大纲"，with strict formatting: use a single-line `---` page break between slides, and every slide must contain exactly two parts: (1) on-slide content (title + concise bullets/short description), (2) a detailed per-slide image description for image generation. |
Ppt Outline
Overview
Turn mixed sources (repo/files/web/context) into a slide-by-slide PPT outline file.
The output is ready for downstream “markdown → ppt” or “markdown → image prompt → ppt” pipelines.
Output Contract (must follow)
- Output: a single Markdown file named
xxx-outline.md.
- If user provides a target path/name, follow it.
- Otherwise default to
outlines/<topic>-ppt-outline-YYYYMMDD.md.
- Slide/page separator:
--- on its own line between slides.
- Every slide must contain exactly two parts:
- 页面内容: what appears on the slide (title + concise bullets / short description)
- 配图描述: detailed image description for image generation
- Keep on-slide text short:
- Prefer 3–6 bullets
- Prefer 6–18 Chinese characters per bullet (or similar brevity in other languages)
- Avoid long paragraphs
- Do not add speaker notes / talk track unless user explicitly requests.
Workflow
Step 0 — Capture parameters (ask only if missing)
Ask up to 3 questions total when unclear:
- Deck goal: persuade / explain / report / training
- Audience: executives / product / engineering / sales / mixed
- Slide count or timebox: e.g., 8/10/12 pages
- Language: zh-CN default
If user already specified constraints, do not ask again.
Step 1 — Collect and prioritize sources
Use only the sources the user points to (do not invent new sources):
- Repo context: read relevant
README.md, docs/, requirements/, workflows/, generated/<name>/workflow.yaml, or files the user mentioned.
- Local files: outlines, inputs, reports, or any referenced docs.
- Web pages: fetch only URLs provided by the user.
- Pasted context: treat as authoritative.
Then extract:
- Key messages (3–6)
- Supporting facts/examples (only those that fit slides)
- Logical story arc (problem → approach → evidence → impact → next steps)
Step 2 — Build the slide storyline
Default storyline (adjust to user intent):
- Title / context
- Problem / opportunity
- Approach / framework
- 2–5 core content slides (group by themes)
- Risks / constraints (optional)
- Summary / next steps
Step 3 — Write each slide with strict template
For each slide:
- Provide a clear slide title
- Write concise bullets that can be pasted into PPT directly
- Write a detailed “配图描述” that matches the slide message and is usable as an image prompt
Step 4 — Quality gate (self-check)
- Every slide has
页面内容 and 配图描述
- Every slide is separated by
---
- No slide is overly verbose
- Image descriptions are specific (subject + scene + composition + style + constraints)
Required Slide Template (copy/paste)
Use this exact structure per slide. Use --- only between slides (do not add a leading separator before slide 1).
Minimal example (2 slides):
1. <标题>
页面内容
配图描述
2. <标题>
页面内容
配图描述
Per-slide template:
<页码>. <标题>
页面内容
- <bullet 1>
- <bullet 2>
- <bullet 3>
配图描述
- 画面主体：
- 场景/环境：
- 构图/镜头：
- 风格/质感：
- 关键细节：
- 画面文字（如需）：
- 避免事项（可选）：
Reference
- If you need a longer example or field guidance for image prompts, read:
references/api_reference.md.
