paper-reading

Produces structured critical reading notes from research paper PDFs with cropped figures and citations.

52|Updated Apr 24, 2026
One-click install
npx skills add https://github.com/Immortalqx/my_codex_skills --skill paper-reading-immortalqx
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: paper-reading
Source: https://github.com/Immortalqx/my_codex_skills/tree/main/paper-reading
Command: npx skills add https://github.com/Immortalqx/my_codex_skills --skill paper-reading-immortalqx

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pymupdf, pillow, and includes scripts (resource) and references (resource) components.

What problem does it solve? Reading a research paper deeply takes hours, and quick summaries miss benchmark misuse, overclaims, and weak evidence. This Skill turns a single paper PDF into a rigorous, citation-backed reading note you can trust months later. ## Core Features & Use Cases - Five-phase deep reading workflow: skim and frame, benchmark audit, prior-art matrix, deep re-read with figure capture, and final note synthesis. - Benchmark audit and critical checks: verifies each benchmark's original task against the paper's usage, flags overclaims, unevaluated scenarios, and reproducibility gaps. - PDF tooling scripts: extract text with page markers, split sections heuristically, render pages to PNG, and auto-crop figures with manual bbox fallback. - Use Case: Hand the Skill an arXiv PDF of a new segmentation paper and receive a 9-section note with cropped figures, a related-work matrix of up to 15 papers, and benchmark-mismatch warnings. ## Quick Start Use the paper-reading skill to deeply read this paper PDF and produce a structured critical reading note with cropped figures and numbered citations.

Frequently Asked Questions about paper-reading

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a structured reading note from a research paper PDF?▼

Provide the PDF, arXiv ID, or URL and ask for a deep reading. The skill runs five phases: framing, benchmark audit, prior-art matrix, deep re-read with figure cropping, and synthesis into a 9-section note saved under readings/<slug>/.

What tools extract text and figures from PDF papers?▼

The bundled scripts use PyMuPDF as the primary backend with Poppler's pdftotext and pdftoppm as fallbacks. Pillow handles figure cropping, and locate_figures.py finds captions automatically with a manual --bbox override for bad crops.

Does this skill work for literature surveys or quick paper overviews?▼

No. It is scoped to deep reading of a single paper. Quick one-paragraph overviews belong to a lookup skill, broad literature surveys to a survey skill, and simulated peer review to a review skill.

Why does figure cropping produce wrong or clipped images?▼

Auto-crop locates the caption and crops above it, which fails with multi-column layouts or unusual caption placement. Re-run locate_figures.py with --bbox "x,y,w,h" in PNG pixels to specify the crop region manually.

Where are downloaded papers and the final note stored?▼

PDFs go into the project's paper directory (papers/ by default), the final note goes to readings/<slug>/ with its figures, and intermediate artifacts stay in x_temp/paper-reading/<slug>/. Nothing is written into the skill folder.