DEV Community

power zhong
power zhong

Posted on

Tried `chuspeeism/dashi-ppt-skill`: A Practical Presentation Skill for AI Agents

#nextjs #showdev #sideprojects #webdev

Tried chuspeeism/dashi-ppt-skill: A Practical Presentation Skill for AI Agents

chuspeeism/dashi-ppt-skill is an AI-agent skill that generates browser-editable presentations from multiple visual themes. The output can be exported to HTML, PDF, and PPTX, which makes it more useful than a simple “generate some slides” prompt.

It is gaining traction quickly—+98 GitHub stars today—because it targets a real workflow gap: turning rough notes, research, or product specs into something a human can still edit in the browser before sharing or presenting.

For an indie builder, the interesting part is the delivery model. Instead of manually wiring a slide-generation service, you can connect the skill to an OpenAI-compatible gateway and keep your model credentials, spend limits, and routing in one place.

Example configuration:

export AI_BASE_URL="https://b-lost.com/v1"
export AI_API_KEY="$B_LOST_API_KEY"
export AI_MODEL="claude-fable-5"
Enter fullscreen mode Exit fullscreen mode

A simple agent-side configuration could look like:

{
  "provider": "openai-compatible",
  "baseURL": "https://b-lost.com/v1",
  "model": "claude-fable-5",
  "temperature": 0.4,
  "headers": {
    "Authorization": "Bearer ${AI_API_KEY}"
  }
}
Enter fullscreen mode Exit fullscreen mode

The architecture is straightforward:

User brief
   ↓
AI agent + dashi-ppt-skill
   ↓
Theme/layout generation
   ↓
Browser editing workspace
   ↓
HTML / PDF / PPTX export
Enter fullscreen mode Exit fullscreen mode

For larger decks, repeated design instructions and brand guidelines can become expensive. If the request path uses Anthropic’s /v1/messages, B-Lost’s native prompt caching may help: cached prompt hits receive a 90% discount, while the relay lists a 20% discount from official pricing.

I’d still test a few representative decks before committing—especially charts, font rendering, and PPTX fidelity. But for fast internal demos, client proposals, and product launches, this looks like a practical skill to add to an AI content pipeline without building a presentation editor from scratch.

Top comments (0)