MD-Skills

Generate publication-quality infographics from any markdown content — talk summaries, research papers, product plans, blog posts, dashboards. Default visual style: modern enterprise AI / SaaS keynote (white background, blue + orange accents, rounded cards, clean sans-serif, generous whitespace). Optional overrides: chalkboard, aged-academia, technical-schematic, pop-laboratory, morandi-journal. Standard outputs: 16:9 landscape infographic per topic/section, 9:16 portrait poster for whole-content overviews, 3:4 portrait for dense-modules infopedia. Trigger when the user asks to "create infographic", "generate infographic", "make a poster", "visualize this content", or provides a markdown file and asks for a visual summary.

Transcribe an audio file (mp3 / wav / m4a) to clean Markdown. Wraps Microsoft's `markitdown` library (via `uvx`); transcription uses the SpeechRecognition library's Google Web Speech API backend (free tier - no API key required for the deterministic step). Two optional summarize sub-modes (host agent does the LLM work, no API key needed) - "single" produces one structured summary MD, "sections" auto-derives 3-8 chapter sections and writes one MD per section + an index (designed to feed downstream chapter infographics). Use when the user provides an audio file or asks to "transcribe this audio", "convert this mp3 to text", "summarize this podcast episode", or "meeting recording to MD".

Convert a Word document (.docx), PowerPoint deck (.pptx), Excel workbook (.xlsx), or CSV file to clean Markdown. Wraps Microsoft's `markitdown` library (via `uvx`) for the deterministic extraction. Two optional summarize sub-modes (host agent does the LLM work, no API key needed) - "single" produces one structured summary MD, "sections" auto-derives 3-8 chapter sections and writes one MD per section + an index (designed to feed downstream chapter infographics). Use when the user provides a `.docx` / `.pptx` / `.xlsx` / `.csv` file path or asks to "convert this Word doc", "extract this presentation", "spreadsheet to markdown", "summarize this report", "summarize this deck", or "break this report into chapter notes".

Convert a PDF file to clean Markdown. Wraps Microsoft's `markitdown` library (via `uvx`) for the deterministic extraction. Two optional summarize sub-modes (host agent does the LLM work, no API key needed) - "single" produces one structured summary MD, "sections" auto-derives 3-8 chapter sections and writes one MD per section + an index (designed to feed downstream chapter infographics). Use when the user provides a `.pdf` file path or asks to "convert this PDF", "extract text from this PDF", "summarize this paper", "give me a TL;DR of this PDF", "break this paper into chapter notes", or "make per-chapter infographics from this PDF".

Convert any URL to clean Markdown. Tries a fast plain-fetch first; if the page is a SPA / login-walled / anti-bot-protected, automatically escalates to headless Chrome. Output goes through Mozilla Readability + Turndown (with GFM) for clean article extraction with no nav/footer/sidebar noise. Two optional summarize sub-modes (host agent does the LLM work, no API key needed) - "single" produces one structured summary MD, "sections" auto-derives 3-8 chapter sections and writes one MD per section + an index (designed to feed downstream chapter infographics). Use when the user pastes any web URL or asks to "save this page as markdown", "convert this URL", "summarize this article", "give me a TL;DR", "break this article into chapter notes", "give me per-section notes", or "make per-chapter infographics from this article".

Fetch a YouTube transcript and (optionally) summarize it into Markdown. Two paths - "transcript" runs a deterministic Bun script that produces a timestamped transcript MD; "summarize" is an agent-driven workflow that turns that transcript into either a single structured summary or multiple per-chapter MDs (designed to feed chapter infographics). NO API key required - all LLM work is performed by the host agent in its own conversation. Use when the user pastes a YouTube URL (youtube.com, youtu.be, shorts) or asks to "get the transcript", "summarize this YouTube video", "convert YouTube to markdown", or "break this video into chapters".