Dashi PPT Skill: Editable Presentation Decks for AI Agents

Key Takeaways#

• Dashi PPT Skill is an agent skill for generating editable browser-based presentations and exporting them to PPTX, PDF, or HTML.
• The README documents 12 visual themes, 1,020 layout pages, and 8,576 adjustable controls.
• It supports Claude Code, Codex, Doubao, Marvis, Workbuddy, Dumate, Qclaw, Cursor, and other local agents that can read files and run shell commands.
• Installation is through npx dashi-ppt-skill@latest, with a China mirror command documented for npm mirror users.

What it is#

Dashi PPT Skill is a practical AI-agent skill for turning source material into editable presentation decks. The project is aimed at workplace slide work: industry research, fundraising recaps, competitor analysis, trend reports, project updates, proposal decks, roadshow materials, and internal training. Instead of asking an agent to produce static slide images, Dashi outputs a browser-editable presentation. Users can change text, swap media, adjust layout, switch colors, and then export the result as HTML, PDF, or a real editable PPTX file.

How it works#

The skill gives an agent a structured presentation system. A user provides documents or requirements, describes the audience and desired point of view, and asks the agent to build a deck. The skill then helps choose among 12 visual themes and uses page roles such as cover, table of contents, charts, comparison pages, process pages, risk pages, and closing slides. Each generated page includes controls for layout, emphasis, module count, colors, and animation.

Installation#

The official README shows:

npx dashi-ppt-skill@latest

For China-based npm mirror users, it also lists:

npx --registry=https://registry.npmmirror.com dashi-ppt-skill@latest

Node.js 20+ and npm are required. Exporting PPTX or PDF also needs Chrome, Chromium, or Edge installed locally.

Why builders should care#

Presentation generation is a common agent demo, but the useful version is editable. Dashi’s design treats HTML as both the agent-readable working format and the user-editable output. That means the agent can generate, inspect, revise, and validate the deck before handoff, while the human can still make final visual edits without recreating slides from scratch. For teams building agent workflows, this is a good example of pairing generated artifacts with a browser-native control surface.

Limits#

The README is clear that Dashi is best for structured business presentations, not pixel-perfect custom art direction. It is a skill and local generation workflow, not a hosted presentation SaaS. Users need a capable local agent environment and the required runtime dependencies.