paperhub-read-paper

name	paperhub-read-paper
description	Read a paper PDF and generate structured three-pass reading notes

PaperHub Read Paper

Use this skill when the user wants to read a paper and create or extend structured reading notes.

What this skill does

• Reads the PDF directly to understand the paper's content
• Generates structured three-pass reading notes in notes/reading-3pass.md
• Supports modes: pass1 (problem/contribution), pass2 (method/tradeoffs), pass3 (reconstruction), full
• Preserves existing notes and appends new structured content

Workflow

1. Identify the target paper. Search first if paper_id is unknown.
2. Read the PDF at library/papers/<paper-id>/pdf/paper.pdf
3. If PDF is not hydrated (local), run:

paperhub organize-paper --workspace <workspace-path> --paper-id <paper-id> --pdf-action hydrate

4. Fill in the three-pass reading structure based on the actual paper content:

Pass 1 - Problem & Contribution:

• Problem: What issue does this paper address?
• Contribution: What does this paper contribute?
• Relevance: Why is this important?

Pass 2 - Method & Analysis:

• Method: How does the paper approach the problem?
• Assumptions: What does it assume?
• Strengths: What works well?
• Weaknesses: What are the limitations?

Pass 3 - Deep Reconstruction:

• Detailed reconstruction: Can you reproduce or trace the key ideas?
• Open questions: What remains unsolved?
• Follow-up experiments: What would you try next?

5. Write notes to notes/reading-3pass.md
6. Return the note path and a summary of key findings.

Operational rules

• YOU must read the PDF — do not just create empty templates. Use the Read tool to read the actual PDF file.
• If the paper has no PDF and no abstract, inform the user and offer to import by URL/title first.
• Do not overwrite user notes; append structured output instead.
• The CLI command paperhub read-paper only creates templates—actual reading is done by you (the AI).