baoyu-slide-deck turns an article into a slide outline, generates an image for each page and combines them as PPTX or PDF. It suits self-contained slides for reading and sharing; the PPTX contains image-based pages.
Install and prepare
Install the skill from your project terminal:
npx skills@latest add jimliu/baoyu-skills --skill baoyu-slide-deck --agent claude-code
Prepare the article or Markdown file, audience and target slide count. The environment needs Bun or Bun through npx, plus a usable image-generation backend. This installs for Claude Code; use --agent codex and $baoyu-slide-deck for Codex.
Your first task
Enter this in the agent conversation, adapting the files and goal to your task:
/baoyu-slide-deck notes/agent-skills.md --style sketch-notes --audience beginners --slides 8 --lang en --outline-only
Explain the point of each slide and how the sequence fits readers who are new to agent skills.
Agree on the outline
For a first attempt, use --outline-only to inspect the structure before generating images. An eight-page example can cover the problem, installation, invocation, one task and common mistakes. Give each slide one main purpose rather than stretching a single point merely to meet a slide count.
Save prompts and choose generation
After agreeing on style, language and audience, per-slide prompts are saved under prompts/. Use --prompts-only to inspect them before creating images. Rendering requires an image backend available to the agent; without one, the outline or prompt stage is still a useful deliverable.
Revise individual pages and merge
Inspect names, numbers and slide titles in the generated images. If one page is wrong, revise its prompt and use --regenerate with the relevant page numbers, such as 3 or 2,5. Merge only after the slides are acceptable and retain the outline, prompts and source images for later edits.
Check before delivery
Open both PDF and PPTX outputs and inspect ordering, sharpness, text accuracy and clipping. Shared slides should remain readable at ordinary viewing sizes, not just look attractive as thumbnails.
Common questions
Why can’t I edit individual text boxes? This is an image-based workflow; baoyu-design is a separate approach.
How should I fix incorrect text? Correct the prompt and regenerate that page to preserve its design.
How many pages generate at once? Batch behavior depends on the backend and configuration; confirm the prompts first.
Turn a training handout into a presentable deck
Suppose a 3,000-word internal handout must become a ten-slide onboarding lesson. State the audience, target count, and what learners should be able to do afterward. Start with --outline-only and check that the pages progress through problem, procedure, demonstration, and verification. If the outline compresses the core action into one slide while repeating background on three others, fix the outline before image prompts. A technical explainer and an executive update may need different information density; a visual preset cannot repair weak structure.
Next use --prompts-only to inspect text and image directions for every slide, especially commands, chart labels, and numbers. Confirm that the selected backend handles references and Chinese text before generation. Use --regenerate on a specific slide with bad wording. Inspect the final deck at projection and mobile sizes for small text and clipping. If recipients need to edit individual words, explain that the PPTX consists largely of whole-slide images and provide the handout, outline, and prompts too. An openable PPTX is not automatically an editable text deck.