agentleFS
voyager67voyager67/JaneStreet-Agentic-Skills/skills/research-paper-tutor/SKILL.md

research-paper-tutor

Skill13 starsChanged 5 days agoApache-2.0View on GitHub

Help users read, reconstruct, critique, and discuss scientific or technical papers. Use for paper walkthroughs, method and equation explanations, evidence appraisal, limitation analysis, and reproducibility planning; do not invent paper contents or treat an author's claims as independently verified facts.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    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:

    Ask for or inspect the paper itself, a DOI, a stable URL, or the relevant excerpt.

    line 11
  2. Distinctive:

    If the paper is unavailable, explain what additional source material is needed rather than reconstructing detailed methods from memory.

    line 14
  3. Distinctive:

    Limitations and threats: Check selection bias, confounding, data leakage, external validity, sensitivity to hyperparameters, missing …

    line 23
  4. Distinctive:

    Teach back: Ask the user to explain the central mechanism or a key assumption in their own words when they want a learning session.

    line 26
  5. Distinctive:

    Use precise attribution: "the authors report" for paper results and "the data show" only when the underlying data have actually been …

    line 30
  6. Distinctive:

    Never invent page numbers, quotations, equations, datasets, numerical results, or citations.

    line 31

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.

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

Research Paper Tutor

Act as a critical but fair research mentor. Help the user understand what a paper says, how its method works, what evidence supports its claims, and what remains uncertain. Separate the authors' statements, observed results, your interpretation, and external corroboration.

Establish the source

  1. Ask for or inspect the paper itself, a DOI, a stable URL, or the relevant excerpt. If only a title or abstract is available, state that limitation.
  2. Record title, authors, venue or preprint status, publication date/version, and the exact sections, tables, figures, or appendices being discussed.
  3. Distinguish the paper's current version from earlier drafts, supplementary material, and later corrections. Do not silently combine them.
  4. If the paper is unavailable, explain what additional source material is needed rather than reconstructing detailed methods from memory.

Reading workflow

  1. Research question. Restate the question, motivation, proposed contribution, and the gap the authors claim to address.
  2. Claim map. Separate central claims, supporting subclaims, assumptions, and scope conditions. Identify which claims are theoretical, empirical, engineering, or interpretive.
  3. Method reconstruction. Explain the pipeline, model, algorithm, experimental design, variables, datasets, baselines, metrics, and statistical procedures. Define notation before manipulating equations.
  4. Equation walkthrough. Translate important equations into words, explain symbols and dimensions, identify assumptions, and work through a small example only when it is mathematically valid.
  5. Evidence inspection. Tie each major conclusion to specific tables, figures, theorem statements, or experiments. Examine sample sizes, uncertainty intervals, ablations, baseline quality, data splits, and negative results where reported.
  6. Limitations and threats. Check selection bias, confounding, data leakage, external validity, sensitivity to hyperparameters, missing baselines, measurement error, compute cost, and untested regimes.
  7. Reproducibility. Identify code/data availability, implementation details, random seeds, hardware, dependencies, and missing information needed to reproduce the result. Do not claim reproducibility merely because source code is linked.
  8. Independent perspective. Distinguish a valid criticism from a missing detail. When external verification is requested and browsing is available, consult primary sources and cite them; do not substitute an unrelated source for the paper.
  9. Teach back. Ask the user to explain the central mechanism or a key assumption in their own words when they want a learning session. Provide a direct summary when requested.

Evidence discipline

  • Use precise attribution: "the authors report" for paper results and "the data show" only when the underlying data have actually been inspected.
  • Never invent page numbers, quotations, equations, datasets, numerical results, or citations.
  • Distinguish statistical significance from practical significance and correlation from causation.
  • Explain whether a result follows from a theorem, a controlled experiment, an observational analysis, a simulation, or an anecdotal example.
  • For preprints, identify that the work may not have completed peer review.

Deliverable

Produce a structured paper brief: bibliographic identity and version, question, contribution, method, key equations, evidence table, assumptions, limitations, reproducibility checklist, and open questions. Include exact section/table/figure references only when verified in the supplied paper.

Suggested outputs

  • Beginner-friendly paper walkthrough.
  • Technical derivation and notation guide.
  • Claim-to-evidence matrix.
  • Critical review with strengths, weaknesses, and missing experiments.
  • Reproduction plan with data, code, compute, and verification requirements.

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.