agentleFS
jxtsejxtse/scientific-research-skills/skills/paper-reading/SKILL.md

paper-reading

Skill68 starsChanged 7 months agoMITView on GitHub

Reads and analyzes academic papers (arXiv preprints, conference / journal PDFs, Zotero items) at three configurable depths: quick skim (2 min), standard read (10 min), or deep analysis (30 min). Produces structured digests covering problem, method, key innovation, results, limitations, reproducibility, hidden assumptions, and connections to the user's other work. Use when the user shares an arXiv link, PDF, or paper title and asks to read / summarize / digest / TL;DR / analyze / review / critique / explain / break down a paper, asks about a paper's contributions / methods / results / equations / figures, wants to compare two papers side by side, or needs a reading note for their records.

20 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 9 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 9

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Follow the appropriate level template above.

    line 76
  2. Distinctive:

    After reading, save the digest:

    line 80
  3. Distinctive:

    If user confirms, add/update Zotero entry with notes

    line 82
  4. Distinctive:

    Check the ablation study — it reveals what actually matters in their method

    line 94
  5. Distinctive:

    Read the limitations/future work section — often more honest than the intro

    line 95
  6. Distinctive:

    Look at Appendix — important details are often buried there

    line 96
  7. Distinctive:

    Check if baselines are fairly implemented (same hyperparameter search budget?)

    line 104
  8. Distinctive:

    Look for cherry-picked examples in qualitative analysis

    line 105
  9. Distinctive:

    When user asks to compare two papers:

    line 109

“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Paper Reading

Structured workflow for reading academic papers efficiently.

When to Use

  • User shares an arXiv link or PDF and asks to read/summarize it
  • User asks about a specific paper's contributions, methods, or results
  • User wants a reading digest for their records
  • User asks to compare a paper against related work

Reading Levels

Level 1: Quick Skim (2 min)

When: User just wants to know if a paper is worth reading

Output:

  • Paper title, authors, venue, year
  • One-paragraph summary (what problem, what method, what result)
  • Key contribution in one sentence
  • Relevance assessment to user's work
  • Recommendation: Read / Skip / Skim only

Level 2: Standard Read (10 min)

When: User wants to understand the paper's approach

Output:

  • Problem: What gap does this address?
  • Method: How do they solve it? (with key technical details)
  • Key innovation: What's genuinely new vs. incremental?
  • Results: Main numbers + comparison to baselines
  • Limitations: What they don't do, acknowledged or not
  • Connections: How does this relate to user's active projects?

Level 3: Deep Analysis (30 min)

When: User is seriously considering building on this paper

Output:

  • Everything from Level 2, plus:
  • Detailed methodology: Step-by-step technical walkthrough
  • Reproducibility assessment: Can you implement this from the paper alone?
  • Experimental design critique: Are the baselines fair? Metrics appropriate?
  • Hidden assumptions: What are they not saying?
  • Extension opportunities: How could this be improved or adapted?
  • Key equations/algorithms: Extracted and explained
  • Figure analysis: What do the key figures actually show?

Workflow

Step 1: Obtain Paper

arXiv link → Download PDF, extract text
PDF file → Extract text directly
Paper title → Search Semantic Scholar → get arXiv link → download
Zotero item → Get from local library

Step 2: Read at Requested Level

Follow the appropriate level template above. When in doubt, start with Level 2.

Step 3: Store Digest

After reading, save the digest:

  1. Store structured summary to local dashboard
  2. If user confirms, add/update Zotero entry with notes

Step 4: Connect to Context

  • Link to user's active projects if relevant
  • Suggest follow-up papers (from references or "cited by")
  • Note if this paper supports or contradicts prior reads

Reading Heuristics

For ML/AI papers:

  • Jump to Table 1 (main results) first — if the numbers aren't impressive, calibrate expectations
  • Check the ablation study — it reveals what actually matters in their method
  • Read the limitations/future work section — often more honest than the intro
  • Look at Appendix — important details are often buried there

For methods papers:

  • Focus on Figure 1 (method overview) + Section 3 (method) + Table 1 (results)
  • Skip related work on first pass — come back only if you need positioning context

For empirical papers:

  • Focus on experimental setup, metrics, and statistical significance
  • Check if baselines are fairly implemented (same hyperparameter search budget?)
  • Look for cherry-picked examples in qualitative analysis

Paper Comparison Mode

When user asks to compare two papers:

| Aspect       | Paper A          | Paper B          |
|--------------|------------------|------------------|
| Problem      |                  |                  |
| Method       |                  |                  |
| Data         |                  |                  |
| Key metric   |                  |                  |
| Advantage    |                  |                  |
| Limitation   |                  |                  |

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.