AI PPT Maker

Problem

Generating PPTX directly from a model often produces inconsistent layouts, unstable cross-slide references, and hard-to-verify icon or font behavior. api_maker2 frames presentation creation as an inspectable pipeline: research, HTML slides, and PPTX export, reducing uncertainty compared with one-shot generation.

How It Works

The skill uses a Research → HTML → PPTX workflow:
- Unified content layer: web research, user-provided documents, or AI-organized material are consolidated into research.md, which downstream outline and slide generation read from.
- Single source of truth: presentation.json stores title, theme, slide order, content_spec, and file mappings, keeping slide additions, deletions, and exports consistent.
- HTML intermediate representation: each slide is an independent HTML file on a 1280×720 canvas, styled with Tailwind CSS and Lucide icons for browser preview and stronger visual control.
- Controlled generation: serial generation is the default; parallel generation is used only when the platform supports sub-agent file writes.

The workflow covers requirement gathering, research collection, outline confirmation, project initialization, slide generation, icon validation, and PPTX export. Icon validation checks Lucide references and attempts to replace invalid names with valid matches.

Boundaries

It fits structured, research-backed decks and office presentations with a clear theme, not open-ended design work. Content quality depends on research.md; export requires Playwright with headless Chromium, and missing fonts can affect layout calculation.