PhDSkills

Initialize, adopt, recover, and govern an AI-first project state document system for any project. Use when Codex must interrogate goals and constraints, scaffold project documents, keep milestones and TODOs recoverable, record acceptance evidence, and preserve failed explorations without lying about progress.

Convert Markdown and Quarto Markdown into DOCX, PDF, and HTML publishing outputs with a DOCX-first workflow. Use when Codex needs to render or validate .md or .qmd files, select or patch a reference.docx template, choose a CSL style, repair non-destructive citation or cross-reference issues, run Quarto or Pandoc publishing commands, or install and verify Quarto, Pandoc, and TinyTeX on Windows or macOS. Prefer the built-in scripts and template assets for deterministic conversion; only use LLM-assisted cleanup when the user explicitly asks for editing or when scripted conversion fails and a fallback is needed.

Turn a report, proposal, thesis chapter, research plan, or other long-form academic text into a highly detailed image-generation prompt for a flat research-overview flowchart. Use when Codex needs to read the source text, extract the internal logical structure between background, objectives, methods, modules, stages, inputs, outputs, and expected results, then produce a Nano Banana compatible prompt that tightly controls layout, wording, aspect ratio, palette, typography, and diagram style. Especially useful for Chinese academic materials such as doctoral proposals, NSFC-style research overview figures, technical roadmaps, and research framework diagrams.

Draft polished AI PhD reports in Markdown from user requirements, outlines, rough notes, existing `.md` text, paper PDFs, or `.bib` files. Use when Codex needs to write or rewrite doctoral proposal reports, daily or weekly or monthly reports, midterm reports, progress summaries, research statements, or other AI PhD-facing academic documents; remove obvious AI-generated tone; preserve a coherent professional narrative; verify factual claims and paper details with live search when needed; and integrate accurate citations or cross-references from user-provided bibliography data or source papers.

Create stunning, animation-rich HTML presentations from scratch or by converting PowerPoint files. Use when the user wants to build a presentation, convert a PPT/PPTX to web, or create slides for a talk or pitch. Helps non-designers discover their aesthetic through visual exploration rather than abstract choices.

Integrate image files into an existing Markdown report by inspecting the actual image content, selecting only the figures that strengthen the report, inserting them at coherent positions, adding captions and cross-references, and renaming the image files into a clean structure. Use when Codex receives a report `.md` plus one or more candidate images and must update the original Markdown in place without rewriting unrelated prose.

Create talk-ready slide blueprints from dense reports, theses, proposals, technical writeups, or long notes. Use when Codex needs to discuss content with the user, compress it into a coherent speaking storyline, strictly limit text per slide, and output a markdown slide design spec covering page-by-page text, layout, visuals, images, tables, contrast, and presentation flow.

Build verified research landscape summaries and curated bibliographies for AI topics. Use when Codex receives a concrete AI research direction, task, or seed papers and needs to search Google Scholar and Google, trace the field through surveys and canonical papers, generate search plans and candidate sets, rank candidates by recency and influence, extract or normalize BibTeX, repair the Python environment when dependencies are missing, and deliver a research-status markdown document plus a source-grounded `.bib` file.