paper-extract
Skill14 starsChanged 4 months agoMITView on GitHub
Use when extracting and analyzing academic papers from PDF or LaTeX sources. Use when the user provides a paper, asks to analyze a paper, understand a paper's contributions, decompose a paper into HLS components, or needs a structured paper breakdown with Overview, Big Background, Small Background, Problems, Root Cause, Key Idea, Design Points, and System evaluation.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- smallSheep123/paper-skillstheir own version, first seen Oct 25, 2026
- lucaswychan/quant-paper-agenttheir own version, first seen Oct 13, 2026
- L4pisLazuli/grounded-paper-skillstheir own version, first seen Aug 19, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- smallSheep123/paper-skills★ 8same name
- L4pisLazuli/grounded-paper-skills★ 0same name
- lucaswychan/quant-paper-agent★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- writing-skillsSkill
- academic-pptx-skillSkill
- analyze-plotSkill
- paper-plotSkill
- paper-reviewSkill
- paper-writingSkill
- systematic-researchSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 32
All E1/E1.5/E2 outputs include OKF YAML frontmatter (type, title, description, tags, status, level, generated, sources).
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Paper Extract — Reverse HLS Decomposition
Overview
Inverse of paper-writing HLS: decomposes a paper into HLS components for deep understanding, gap identification, and research positioning. Outputs in OKF format (YAML frontmatter + structured Markdown) for knowledge graph integration.
Templates → templates/. OKF format → references/okf-format.md.
When to Use
Decision flowchart → references/decision-flow.md.
Use when: user provides PDF/LaTeX to analyze; literature review; pre-review deep read; gap identification; paper comparison.
Don't use for: writing → paper-writing; reviewer feedback → paper-review; figures → paper-plot.
Process Flow
| Phase | Role | Output | Template |
| E0 | Extractor — text + figures from PDF/LaTeX | output/analysis/ | — |
| E1 | Decomposer — HLS 7+1 core points with evidence | l0.md (OKF) | templates/l0.md |
| E1.5 | Condenser — TLDR, Abstract+Intro only | tldr.md (OKF) | templates/tldr.md |
| E2 | Architect — Section flow chains | l1.md (OKF) | templates/l1.md |
| E3 | Critic — innovation, gaps, positioning | output/analysis/critical-assessment.md | — |
All E1/E1.5/E2 outputs include OKF YAML frontmatter (type, title, description, tags, status, level, generated, sources).
Output layout → references/extraction-guide.md.
Hard Gates
<HARD-GATE-E0>
Full text + figure PNGs + captions + references extracted. Verify `output/analysis/full-text.md` and `figures/` exist.
</HARD-GATE-E0>
l0.md` with OKF frontmatter and evidence citations. Descriptive, not evaluative. Overview ≤5 sentences. Key Idea = 1 sentence.
Template → [templates/l0.md](templates/l0.md).
</HARD-GATE-E1><HARD-GATE-E1.5>
TLDR from Abstract+Intro ONLY. OKF frontmatter REQUIRED. Structured (1)(2)(3) sub-points, 1-2 sentences each. No [MISSING].
Template → templates/tldr.md.
</HARD-GATE-E1.5>
<HARD-GATE-E2>
Each Section's A→B→C chain + figure placements + mermaid dependency graph. OKF frontmatter REQUIRED.
Template → [templates/l1.md](templates/l1.md)
<HARD-GATE-E2>
Each Section's A→B→C chain + figure placements + dependency graph. User confirms.
</HARD-GATE-E2><HARD-GATE-E3>
Innovation level + 3 strengths + 3 weaknesses + gap analysis + positioning map. Every claim cites paper text.
</HARD-GATE-E3>E1: The 7+1 Core Points
| # | Point | Question Answered | Paper Section Clue |
| 1 | Overview | What does this paper do in one paragraph? | Abstract + Introduction |
| 2 | Big Background | What broad research wave/trend enables this? | Introduction ¶1-2 |
| 3 | Small Background | What specific prior work does this directly extend? | Introduction ¶3-4, Related Work |
| 4 | Problems | What concrete problems does this paper solve? | Introduction (gap statements), Problem Formulation |
| 5 | Root Cause | What fundamental bottleneck underlies all problems? | Motivation, Design Rationale |
| 6 | Key Idea | What is the ONE core insight? (≤1 sentence) | Introduction (thesis statement), Contribution list |
| 7 | Design Points | What are the 2-5 technical contributions? | Method/Design/Architecture sections |
| +1 | System & Experiment | How was it implemented and evaluated? | Implementation, Evaluation |
For each point: state claim → cite evidence (quote or section:paragraph) → flag [MISSING] if no evidence found.
E1.5: TLDR
Structured (1)(2)(3) sub-points per section, 1-2 sentences each. Abstract+Intro ONLY. Template → references/tldr-format.md.
E2: Section Flow Chains
For each Section: name, purpose, A→B→C chain, figure refs to output/analysis/figures/, dependencies. Dependency graph → references/extraction-guide.md.
E3: Critical Assessment
Innovation level (Breakthrough/Novel/Incremental/Engineering), 3+ strengths/weaknesses, gap analysis, positioning map (approach type × scope, 3-5 prior works). Details → references/extraction-guide.md.
Quick Reference
| User says | Trigger mode |
| "Analyze this paper" | E0→E1→E1.5→E2→E3 |
| "Give me a quick summary" | E0→E1→E1.5 |
| "Just the TLDR" | E0→E1.5 |
| "What's the core idea?" | E0→E1 (Key Idea + Design Points) |
| "Find the gaps" | E0→E1→E3 (Gap Analysis) |
| "Compare these two papers" | Full pipeline on both, side-by-side |
| "Is this paper good?" | E0→E1→E3 (Strengths/Weaknesses) |
Source Type Handling
PDF: scripts/extract-pdf.py <paper.pdf> --output output
LaTeX: scripts/extract-latex.py <main.tex> --output output
Details → references/extraction-guide.md. Run scripts with --help for options.
Common Mistakes
| Mistake | Fix |
| Summarizing instead of decomposing | Each HLS point is a distinct dimension |
| Evaluating during E1 | E1 = descriptive only. Judgment → E3 |
| Fabricating missing points | Mark [MISSING] — don't invent |
| Confusing Problems with Root Cause | Problems = what's broken; Root Cause = WHY |
| Skipping evidence citations | Every hls-L0.md point MUST cite paper text |
| Relying on captions alone | Extract figure images to figures/ for visual inspection |
| TLDR using full paper | TLDR = Abstract+Intro ONLY |
| TLDR as single blob paragraph | Must use (1)(2)(3) line breaks — never a wall of text |
Cross-Skill Integration
| Skill | Relationship |
| paper-writing | Inverse: paper→HLS vs HLS→paper |
| paper-review | E3 feeds into review dimensions |
| analyze-plot | E2 figure maps guide regeneration |
REQUIRED BACKGROUND: paper-writing's HLS framework (L0 core points, L1 flow chains).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.