agentsclimarketplace

Learn from paper

Skill leon2k2k2k/agent-skills/learn-from-paper

Production-tested skills for AI coding agents. Compatible with Claude Code, Codex, Gemini CLI, Cursor, and more.

Install
npx -y skills add leon2k2k2k/agent-skills --skill learn-from-paper

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Interactive paper reading session. Walk through a paper section by section with Q&A, quizzes, testing, and note-taking.

The file declares its own license as Apache-2.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

4.5 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it

Reading Session

Interactive walkthrough of a research paper. The user drives the pace; the agent explains, answers questions, and takes notes.

Phase 1: Get the paper

The user can provide the paper in any of these ways:

  1. arXiv URL (e.g. https://arxiv.org/abs/1706.03762) — download the TeX source from https://arxiv.org/src/{arxiv_id}, cache to ~/.cache/papers/{arxiv_id}/, find the entrypoint .tex file
  2. Local PDF (e.g. ./paper.pdf) — read the PDF directly
  3. Local TeX file (e.g. ./paper.tex) — read the TeX source directly
  4. Paper name (e.g. "Attention Is All You Need") — search for it, ask the user to confirm, then fetch

If no argument is given, ask: "Do you have an arXiv link, a local file (PDF or TeX), or a paper name?"

Phase 2: Overview & story

  1. Tell the story. Don't just summarize — narrate the problem the paper solves, the history leading up to it, why existing approaches failed, and what breakthrough this paper represents. Make it compelling.
  2. List the paper's sections so the user can see the structure
  3. Ask the user where they want to start, or suggest going section by section

Phase 3: Interactive walkthrough

This is the core loop. For each section or topic:

  1. Summarize the section in plain language — focus on intuition, not just restating the text
  2. Highlight the key equations or concepts. For a technical reader: connect to familiar ideas (e.g. "attention is a soft dictionary lookup", "this is just a variational bound")
  3. Prioritize — when moving to a new section, tell the user what's must-know vs nice-to-know vs skippable for their goals
  4. Check understanding — after explaining a concept, check if the user has questions or if something was unclear. Don't just move on.
  5. Wait for the user — let them ask questions, challenge claims, or move on
  6. When the user asks a question, answer concisely. Use the paper's own notation. Reference specific sections/equations.
  7. Do NOT move to the next section until the user says to
  8. Cover everything — after finishing a topic, mention what's left and ask if they want to continue. Make sure nothing important is skipped.

Guidelines

  • Keep explanations concise — adapt to the user's level
  • Don't over-explain basics unless asked
  • If the user asks about connections to other papers, make those connections
  • Be honest about limitations or controversial claims in the paper
  • Never suggest wrapping up. The user decides when they're done. Only finalize notes when the user explicitly says to.

Phase 4: Draft notes (ongoing)

After the overview (Phase 2), immediately create a draft notes file with the template structure below. Update it as the session progresses — add to "What Clicked," "What Was Tricky," "Discussion Notes" in real time. This way nothing is lost and the user can see progress.

Phase 5: Quiz

The user can request a quiz at any time during the session. Also, before wrapping up, always ask if they want one.

  • 5-8 questions over the paper's key concepts. Mix question types:
    • Conceptual: "Why does the Transformer use multi-head attention instead of a single attention function?"
    • Technical: "What is the computational complexity of self-attention vs. a recurrent layer for sequence length n?"
    • Connection: "How does this relate to [concept from user's background]?"
  • Ask one question at a time. After each answer, give feedback — confirm what's right, gently correct what's off, and add nuance.
  • If the user gets something wrong or wants to revisit, go back to that topic in the paper before continuing.

Phase 6: Notes

When the user is done (or explicitly says to finalize notes):

  1. Polish the draft notes file
  2. Structure:
    # Paper Title — Authors, Year
    
    ## Key Ideas
    - bullet points of the main contributions
    
    ## Architecture / Method
    - how it works, with key equations if relevant
    
    ## What Clicked
    - concepts that were easy or intuitive (and why)
    
    ## What Was Tricky
    - parts that needed more discussion or were initially confusing
    
    ## Discussion Notes
    - interesting questions that came up, back-and-forth highlights, insights from the session
    
    ## Connections
    - links to other papers or concepts
    
  3. Ask the user where to save the notes and if they want to commit

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most learn study skills give in ~1.0k tokens

Counted across 546 of the 573 authors here whose files we hold, read 2026-08-07

  • Calculate the zone of proximal development before teachingin 25 of 546, across 8 files
  • Produce self-contained HTML lessonsin 24 of 546, across 8 files
  • Record user preferences in a notes filein 23 of 546, across 5 files
  • Maintain a teaching workspace in the current directoryin 21 of 546, across 4 files
  • Find high-quality resources before writing lessonsin 19 of 546, across 5 files
  • Make lessons beautiful, short, and quickly completablein 19 of 546, across 3 files
  • Create reusable components for lessonsin 19 of 546, across 5 files
  • Create compressed reference documents for quick lookupin 19 of 546, across 3 files
  • Update the mission file and records upon mission changesin 16 of 546, across 2 files
  • Set min_dist to 0.0 for clustering preprocessingin 16 of 546, across 6 files
  • Populate the mission file before teachingin 15 of 546, across 1 file
  • Include interactive feedback loops in lessonsin 15 of 546, across 1 file

Said here and by no other author read

  • fetch the paper from the provided source
  • narrate the problem the paper solves
  • list the paper sections for the user
  • summarize each section in plain language
  • use the paper notation in answers
  • maintain a draft notes file during the session

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.