# dashi-ppt-skill: an AI agent skill that turns documents into browser-editable slide decks

> Dashi PPT Skill installs through npm and hands your agent a library of 12 themes and 1020 layouts. The deck it produces is itself an editor, and it exports to HTML, PDF and editable PPTX. Here is how the pieces fit and where it stops.

**chuspeeism/dashi-ppt-skill** — An AI-agent skill that generates browser-editable presentations from multiple visual themes, exportable to HTML, PDF, and PPTX.

- Repository: https://github.com/chuspeeism/dashi-ppt-skill
- Stars: 8,977 · Forks: 806
- Language: JavaScript
- License: AGPL-3.0
- Published: 2026-09-10 · Updated: 2026-09-10 · Language: en
- Canonical page: https://hysenlabs.com/projects/chuspeeism-dashi-ppt-skill

## What dashi-ppt-skill is for, and who it is not for

The problem this skill addresses is the gap between a document and a presentation. An agent can summarize a report, but a summary is not a deck: pages need roles, a visual language, and room for the presenter to change their mind afterwards. dashi-ppt-skill is an agent skill, meaning it is a set of instructions and assets an AI agent loads, that takes source material and produces a multi-page presentation. The README describes the target user plainly as people who need a deck that is structurally complete, visually consistent, and still editable.

The stated fit is industry research, fundraising retrospectives, competitive analysis, trend reports, project updates, proposals, roadshow material and internal training. The stated mismatch is equally direct: scenarios that need pixel-by-pixel manual visual customization. That is an honest boundary. If your brand guidelines specify exact coordinates for a logo lockup on every page, a generator that assembles pages from a layout library will fight you.

## Layouts, themes and the control console: how a page is actually built

According to the README, the skill ships 12 visual themes and 1020 layout pages, with 8576 adjustable controls across them. Each theme has its own page structures and visual language, and the library covers 20 page roles including cover, table of contents, metrics, trends, comparison, process, risk and closing. Chart layouts include radar, waterfall, treemap, funnel, heatmap, Sankey and Gantt, and there are analysis-model layouts for SWOT, Porter's Five Forces, PEST, business model canvas and double diamond.

The mechanism that matters is that a page is described as a combination of a layout plus copy fields. That is why the output is editable in a browser: the deck is HTML where each page carries its own console with sliders, toggles and dropdowns. The README lists more than 20 dimensions of adjustment per page, covering content, layout, module count, page emphasis, preset color schemes and page transitions. Text is edited in place by clicking, and media slots are replaced by click or drag, with uploads compressed automatically. The README also notes that when your source is text only, the skill reserves image placeholders you can fill later.

The trade-off is visible in that design. A layout-plus-fields model buys you fast, consistent pages and cheap restyling, because the agent can swap a theme across the whole deck. It also means every page is constrained to the layouts the library provides. The README's own framing, that editing after generation matters more than generation, is a fair description of where the effort sits.

## Installing dashi-ppt-skill and producing a first deck

The README gives a single npm command for both install and update, and says rerunning it updates in place while keeping already installed dependencies. The documented environment requirement is Node.js 20 or newer with npm, and a local Chrome, Chromium or Edge for PPTX and PDF export.

```bash
npx dashi-ppt-skill@latest
```

For networks where the default registry is slow, the README gives a mirror variant. It is the same install with a different registry, not a different package.

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

You can also delegate the install to the agent itself. The README supplies this prompt, which you paste into the agent you intend to use.

```text
帮我安装 skill：npx dashi-ppt-skill@latest，国内镜像 npx --registry=https://registry.npmmirror.com dashi-ppt-skill@latest
```

The repository layout shows a skills/ directory and a .claude-plugin/ directory at the top level, which is consistent with the README's claim that for tools like Marvis, Workbuddy, Dumate and Qclaw you can place the skill files anywhere as long as the agent reads SKILL.md. After installation, the documented workflow is five steps: describe the requirement (topic, audience, page count, the conclusion you want to land), pick one of the 12 theme previews and confirm whether you need images or video, let the skill organize the material into structured content and a deck plan, edit in the browser, then export. The README says this takes a few minutes.

## Export paths and the local browser dependency

The README lists three export targets: an offline HTML package, PDF, and editable PPTX. The PPTX export is described as restoring nodes one by one with text remaining editable, and the repository includes a comparison GIF of the HTML version against the exported PPTX page by page. Browser-side editing includes nine page transition animations, a presentation mode, a light and dark theme switch, and a reset for all changes. The thumbnail rail on the left supports drag reordering, skipping, deleting and duplicating pages.

The constraint to plan around is the browser requirement. PDF and PPTX export need Chrome, Chromium or Edge installed on the machine, so a headless container without one of those binaries will generate decks but not export them. The README does not document what happens when the browser is missing, and it does not describe a fallback renderer. Treat the browser as a hard prerequisite for the export step rather than an optional extra.

## Where it is the wrong tool, and what the README leaves open

The clearest failure case is stated by the project itself: pixel-level manual visual customization. A layout library gives you consistency across pages, not control over a single page's composition.

The second case is agent capability. The README's platform table marks plain web chatbots as not recommended, because the generator needs a local Node.js environment. A hosted chatbot that cannot read and write files or run shell commands cannot drive this skill, no matter how good its text output is. The table also notes that Cursor and other local agents work if they can read and write files and execute shell commands, and that Doubao requires office mode to be enabled. Codex is listed as able to call image generation to supplement illustrations.

Several things are simply not documented. The README does not explain how to uninstall the skill or roll back an update, and it does not describe a versioning or migration policy for decks generated under an older theme set. The only release in the repository is readme-assets-v1, whose own title says it is README demonstration material for image hosting and not a software version, so there is no published changelog to consult. The README also does not specify how a custom theme would be added, which matters if the 12 built-in themes do not match your brand.

## How this differs from python-pptx and from template-driven generators

The obvious alternative for programmatic decks is python-pptx, which writes OOXML directly. The difference in approach is where the editing happens. python-pptx produces a file that you open in PowerPoint or Keynote; there is no browser editor, no per-page console, and no theme swap after generation. Changing the visual language of a finished deck usually means regenerating it or editing the file by hand. dashi-ppt-skill inverts that: the HTML deck is the working artifact, and PPTX is an export target.

The other comparison is template-driven generators, where you fill placeholders in a fixed .pptx template. Those give you exact control over the template's look and nothing else. dashi-ppt-skill instead gives the agent 1020 layouts to choose from and gives you sliders to change module counts and page emphasis afterwards. More flexibility during authoring, less determinism about which layout a given paragraph lands in. If your workflow depends on a fixed corporate template, a placeholder filler is the better fit.

## Licence and the cost of keeping it current

The repository is licensed AGPL-3.0. That is a strong copyleft licence with a network clause: if you modify the software and let users interact with it over a network, the licence's terms attach to the modified source. For an internal deck generator on a laptop this is unlikely to raise questions. For anyone embedding the skill in a hosted product, the scope of that clause is the thing to read before shipping. This is a description of the licence, not legal advice.

On maintenance, the last push to the repository was on 2026-09-08, and the repository is not archived. The upgrade path documented in the README is the install command itself: rerunning npx dashi-ppt-skill@latest updates in place and preserves installed dependencies. There is no separate update command and no documented way to pin a version for reproducibility, which is worth knowing if you generate decks in CI.

## Conclusion

Adopt dashi-ppt-skill if you already drive a local coding agent such as Claude Code, Codex or a file-and-shell capable tool, and your deliverable is a structured deck you will keep editing rather than a poster. Skip it if you need pixel-level art direction, or if your only agent is a hosted web chatbot with no local Node.js, since the README lists that case as not recommended. Before committing, verify that Node.js 20 or newer and Chrome, Chromium or Edge are present on the machine, because PDF and PPTX export depend on a local browser, and read the AGPL-3.0 LICENSE in the repository to decide whether its network-copyleft scope fits how you plan to distribute the output.

## FAQ

### What is dashi-ppt-skill?

It is an AI-agent skill that generates browser-editable presentations from source documents, using 12 visual themes and 1020 layout pages. The result can be exported as an offline HTML package, PDF, or editable PPTX.

### How do I install dashi-ppt-skill?

The README gives one command, npx dashi-ppt-skill@latest, and notes that rerunning it updates in place while keeping installed dependencies. It requires Node.js 20 or newer and npm, plus a local Chrome, Chromium or Edge for PDF and PPTX export.

### Can dashi-ppt-skill export PDF or editable PPTX?

Yes, the README lists HTML, PDF and editable PPTX as export targets, with PPTX described as restoring nodes one by one while keeping text editable. PDF and PPTX export require Chrome, Chromium or Edge installed on the machine.

## Sources

- [chuspeeism/dashi-ppt-skill on GitHub](https://github.com/chuspeeism/dashi-ppt-skill)
- [Issues](https://github.com/chuspeeism/dashi-ppt-skill/issues)
- [License: AGPL-3.0](https://github.com/chuspeeism/dashi-ppt-skill/blob/main/LICENSE)
- [README](https://github.com/chuspeeism/dashi-ppt-skill/blob/main/README.md)
- [Releases](https://github.com/chuspeeism/dashi-ppt-skill/releases)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/chuspeeism-dashi-ppt-skill
