phd-deepread

PhD Deep Read Workflow

Transform academic PDFs into structured literature notes and critical-thinking canvases for Obsidian.

---

"The human mind... operates by association. With one item in its grasp, it snaps instantly to the next that is suggested by the association of thoughts." — Vannevar Bush, As We May Think, 1945

Bush imagined this in 1945 — a desk where all your reading lives outside its original containers, ready for you to build trails of association through it. That's what this workflow is.

Step one: Convert the PDF to Markdown. The knowledge leaves the frozen PDF and becomes something you can return to, search, and connect — today, next month, next year.

Step two: Come back anytime and draw out what's relevant now. Connect ideas across papers. Visualize arguments, evidence, and assumptions side by side in an Obsidian canvas. The literature note captures what's there. The canvas lets you see it alongside everything else you know.

---

Prerequisites

You need both of these:

• A skill-capable AI coding agent — either Claude Code (download here) or Codex. This workflow installs as a skill in whichever one you use.
• Obsidian — download here. The final output is a JSON canvas file that only renders in Obsidian.

If you just want AI help reading a PDF, you don't need this workflow. Drag the PDF into any AI chat and ask questions directly. This workflow is for people who want structured literature notes and a critical-thinking canvas they can revisit in Obsidian.

---

Install in 30 seconds

Paste this into Claude Code or Codex:

Install this skill for me: https://github.com/heleninsights-dot/phd-deepread-workflow

Your agent installs the skill, the Python CLI, and everything else automatically. The skill lands in that agent's skills folder (~/.claude/skills/ for Claude Code, ~/.codex/skills/ for Codex) — restart the agent if it asks you to. Done. Drag a PDF in and say "phd-deepread read this paper".

Optional: Tesseract OCR

Only needed if you work with scanned PDFs (image-based, text not selectable):

brew install tesseract          # macOS
sudo apt install tesseract-ocr  # Ubuntu/Debian

---

Using the workflow

One paper

Drag a PDF into Claude Code or Codex and say:

phd-deepread read this paper

Your agent extracts the text, writes a structured literature note, and creates a 9-node critical-thinking canvas — all in one go. Open the .canvas file in Obsidian to visualize arguments, evidence, assumptions, and gaps side by side.

You can also ask for specific parts: "phd-deepread extract this PDF, but just give me the prompt — I'll write the note myself."

A folder of papers

Drag a folder into Claude Code or Codex and say:

phd-deepread read this folder

Your agent batch-processes every PDF inside — extracting text, writing a structured literature note for each one, and creating canvas templates. Already-processed papers are skipped automatically. Works the same as a single paper, just for the whole folder.

---

What you get

Each PDF you process gives you three files:

Output	What it is
paper.md	Full text of the PDF, converted to Markdown
paper_literature_note.md	Structured academic note — see below for what's extracted
paper.canvas	A 9-node critical-thinking canvas — open this in Obsidian to visualize arguments, evidence, assumptions, and gaps side by side

What the literature note extracts

The note doesn't summarize — it pulls out the specific data you need to actually use the paper in your research: