A Python PPT-generation skill for AI coding tools with Build/FreeStyle/VI modes, 40k+ styles and AI imagery — not a DeepSeek Harness plugin.
DSH integration
Ecosystem-related
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-22
License
MIT
01What can it help you accomplish?
Generate a fully editable .pptx from a natural-language prompt inside an AI coding tool
A real .pptx produced through one of three engines — Build Script (per-page pixel-perfect control via the `build_helpers` toolbox), FreeStyle (one-line prompt or a `content.json` with per-slide goals), or VI Build (enterprise VI-compliant generation from a template) — so the deck stays editable in PowerPoint afterwards
Designers and developers who build slides in AI coding tools (OpenCode, Claude Code, Codex, Cursor, …) and want deterministic, editable output rather than a flattened image
Use a large, controllable style system without hand-designing every slide
40,000+ style combinations (30 color palettes × 25 font pairings × 15 decorations × 12 layouts) plus 41 mood keywords and a built-in design database (192 palettes · 84 styles · 74 font pairings · 161 anti-patterns, BM25 searchable); name a preset (e.g. `dark-tech`) or a mood for deterministic output
Users who need brand-consistent or richly styled decks fast, with reproducible results across runs
Add AI-generated imagery and vector graphics to slides
AI image engines (Seedream, GPT Image, DALL-E, Gemini, Wanx + Kimi enhancement, cache-first) plus an SVG→PPTX compiler that turns SVG vectors into native editable shapes, 10 diagram engines and a 12+ transition / 8 entrance / 8 exit / 8 emphasis animation system
Users who want illustrations, charts and motion without leaving the PPTX format
02How to install into DeepSeek Harness
Prerequisites
- Python 3.10+ (badge in README: 'Python 3.10+')
- git, and an AI coding tool from the 13 documented platforms (OpenCode, Claude Code, Codex, Cursor, Windsurf, Roo Code, Gemini, Trae, Continue, Droid, KiloCode, Augment, Copilot). DeepSeek Harness is NOT a documented platform
- `python-pptx` 1.0.2 (the package pulls it in; shown as a dependency badge)
Installation steps
- 01
Clone and run the installer: `git clone https://github.com/sunchaokun/PPT-Design-Skill.git && cd PPT-Design-Skill && python install.py` (auto-detects platform; use `python install.py --platform opencode` to specify, or `--force` to overwrite)
$ git clone https://github.com/sunchaokun/PPT-Design-Skill.git && cd PPT-Design-Skill && python install.py
- 02
Or install as a Python package: `pip install --upgrade git+https://github.com/sunchaokun/PPT-Design-Skill.git` — after this, `ppt_pro_max` is importable from any project
- 03
In your AI coding tool, invoke it with a prompt such as `用Build模式生成一份AI融资路演PPT` and the skill generates the .pptx
Verify the integration
Not specified by the author
03DSH integration and capability boundaries
PPT Design Skill is a SKILL / Python package for AI coding tools, not a DeepSeek Harness (`dsh`) plugin. Its README documents 13 target platforms — OpenCode, Claude Code, Codex, Cursor, Windsurf, Roo Code, Gemini, Trae, Continue, Droid, KiloCode, Augment and Copilot — and DeepSeek Harness is NOT among them, so there is no documented native dsh integration. It can be used alongside dsh only if your dsh exposes a generic skill/agent mechanism, which the README does not describe.
Build Script mode (recommended)
a hand-written `build.py` using the `build_helpers` toolbox→a pixel-perfect, fully editable .pptx — `rect`/`text`/`gradient_text`/`kpi_card`/`native_chart`/`donut_chart`/`code_block`/`section_divider`/… with dark-theme support and `clean_save` that fixes PPTX validity
FreeStyle mode
a one-line prompt (`ppt-design "AI产品融资路演"`) or a `content.json` with per-slide goals→a deck through 11 goal layouts (hook/problem/features/data/code/exercise/diagram/section/testimonials/cta/content) with themed charts, cards and graphics; preset names give deterministic output
VI Build mode (enterprise)
an enterprise `template.pptx`→a VI-compliant deck: `template_analyzer` produces an analysis the LLM turns into `build.py`, preserving the framework slides and build helpers
40,000+ style system & AI image generation
a natural-language `--style` (e.g. `dark cyberpunk`), optional `--fetch-images --llm-provider seedream`→30×25×15×12 style combinations, 41 mood keywords and a BM25 design database; AI imagery from Seedream/GPT Image/DALL-E/Gemini/Wanx with cache-first reuse
Native python-pptx output
design intent from any of the three modes→a fully editable .pptx (not an HTML→screenshot), with 10 diagram engines (flow/funnel/timeline/SWOT/matrix/cycle/table/hierarchy/pyramid/venn), CJK font fallback and an animation system
04Who is it for? When not to use it?
Good for
- Designers and developers who build slides in AI coding tools (OpenCode, Claude Code, Codex, Cursor, …) and want deterministic, editable output rather than a flattened image
- Users who need brand-consistent or richly styled decks fast, with reproducible results across runs
- Users who want illustrations, charts and motion without leaving the PPTX format
Not for
- Not wired into DeepSeek Harness: the README lists 13 target platforms (OpenCode, Claude Code, Codex, Cursor, Windsurf, Roo Code, Gemini, Trae, Continue, Droid, KiloCode, Augment, Copilot) and DeepSeek Harness is NOT among them. There is no documented `dsh plugin` install path, so using it with dsh depends on dsh exposing a generic skill/agent mechanism the README does not describe.
05Compatibility, maintenance and safety notes
- Not wired into DeepSeek Harness: the README lists 13 target platforms (OpenCode, Claude Code, Codex, Cursor, Windsurf, Roo Code, Gemini, Trae, Continue, Droid, KiloCode, Augment, Copilot) and DeepSeek Harness is NOT among them. There is no documented `dsh plugin` install path, so using it with dsh depends on dsh exposing a generic skill/agent mechanism the README does not describe.
- Requires Python 3.10+ and `python-pptx` 1.0.2 (pulled in as a dependency). Install is via git clone + `python install.py`, or `pip install --upgrade git+https://github.com/sunchaokun/PPT-Design-Skill.git`; the README documents update (`git pull && python install.py --force`) but no uninstall/rollback command.
- AI image generation needs an external provider and key (e.g. `--llm-provider seedream --llm-api-key $ARK_API_KEY` for Seedream, or gpt-image/dalle/gemini/wanx). Without a key the default placeholder/search modes are used; this is a per-feature requirement, not for the core deck generation.
MIT · Python package / SKILL for AI coding tools — no DeepSeek Harness plugin documented in README
06Frequently asked questions
What is PPT Design Skill and how is it installed?
It is a SKILL / Python package (MIT). Install by cloning and running `python install.py` (auto-detects your AI coding tool), or `pip install --upgrade git+https://github.com/sunchaokun/PPT-Design-Skill.git`. After install, `ppt_pro_max` is a global Python package and you can prompt your tool with e.g. `用Build模式生成一份AI融资路演PPT`.
Does it work with DeepSeek Harness?
Not as documented: the README lists 13 platforms (OpenCode, Claude Code, Codex, Cursor, Windsurf, Roo Code, Gemini, Trae, Continue, Droid, KiloCode, Augment, Copilot) and DeepSeek Harness is not among them. There is no `dsh plugin` install path described, so any dsh usage would rely on dsh exposing a generic skill mechanism the README does not cover.
What are the three modes?
Build Script (recommended) gives pixel-perfect per-page control via `build_helpers`; FreeStyle takes a one-line prompt or a `content.json` with 11 goal layouts; VI Build generates enterprise VI-compliant decks from a `template.pptx`. Recommended flow: FreeStyle for a quick prototype, then Build for the delivery-grade deck.
How large is the style system and is output deterministic?
30 color palettes × 25 fonts × 15 decorations × 12 layouts = 40,000+ combos, plus 41 mood keywords and a built-in design database. Natural-language styles pick a random palette/font when no seed is given; use a preset name (e.g. `dark-tech`) or `--style-seed` for reproducible output.
How does AI image generation work, and is it required?
AI imagery uses Seedream / GPT Image / DALL-E / Gemini / Wanx (with Kimi enhancement) via flags like `--fetch-images --llm-provider seedream --llm-api-key $ARK_API_KEY`; results are cache-first. It is optional — without a key the default placeholder/search image modes are used and core deck generation still works.
07Related DSH workflows
openviking
by volcengine
Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
colleague-skill
by titanwings
Turn memories and knowledge into warm, reusable agent skills — a skill generator for digital companions in the AI persona era.
nocobase
by nocobase
NocoBase is an open-source AI + no-code platform for building business systems fast. Instead of generating everything from scratch, AI works on top of production-proven infrastructure and a WYSIWYG no-code interface, so you get both speed and reliability.
archify
by tt-a1i
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
08Data and sources
支持 13 个平台：OpenCode · Claude Code · Codex · Cursor · Windsurf · Roo Code · Gemini · Trae · Continue · Droid · KiloCode ·…
在 AI 编码工具中输入：`用Build模式生成一份AI融资路演PPT` → 自动调用 skill → 输出可编辑 .pptx
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-22. Found an error? Submit a correction.