agentleFS
mkaremm80mkaremm80/My-Skills-on-Claude/skills/paper-equation-guide/SKILL.md

paper-equation-guide

Skill52 starsChanged 12 days agoMITView on GitHub

Explain every equation of a research paper in a colored PDF guide: labeled equation terms, plain meaning, worked numeric examples, helper drawings and simulation checks. Use when asked to explain a paper's math.

  • Installs packages

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 1

    1. mkaremm80/-my-claude-skills★ 1100% same

    Edited · 0

    None found.

    Wrote their own version · 0

    None found.

    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 repository5 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 3 are distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read the whole paper. List every numbered equation, proposition/lemma/theorem, the key unnumbered formulas (cost/complexity, metrics such …

    line 13
  2. Distinctive:

    Look for a teaching example that shows WHY an equation is needed (e.g., two cases with the same mean but different outage.

    line 19
  3. Distinctive:

    Verify every closed-form result by Monte Carlo (about 1e6 samples) or direct computation, and record formula vs.

    line 20
  4. Very specific:

    Do NOT name a Python script numbers.py (it shadows the standard library module and breaks numpy).

    line 21
  5. Very specific:

    Add notes with ax.annotate(..., arrowprops=dict(arrowstyle='-|>', color='#C0392B'), bbox=dict(boxstyle='round,pad=0.3', fc='#FFF8E1', …

    line 29
  6. Very specific:

    Use the LaTeX package annotate-equations (needs two or more pdflatex runs):

    line 33
  7. Distinctive:

    Keep labels short (3-6 words).

    line 50
  8. Very specific:

    Use \clearpage before major sections that hold big boxes and \Needspace{12\baselineskip} before smaller headings to avoid orphan headings.

    line 65

≈ “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

Paper Equation Guide

Turn the mathematics of any research paper (PDF or LaTeX) into a colored, easy-to-read PDF guide. Every equation is shown with each term highlighted in color and labeled with an arrow, then explained in plain words, derived briefly, illustrated with real numbers and (when useful) a drawing, and checked by simulation.

Writing style: English, short sentences, simple words (B2 level), professional. Numbers quoted from the paper must match the paper. Numbers from your own worked example must be clearly labeled as an illustration (state the parameters and that one example realization or fixed seed is used).

Step 1 - Read and map the math

  1. Read the whole paper. List every numbered equation, proposition/lemma/theorem, the key unnumbered formulas (cost/complexity, metrics such as reduction %, confidence intervals, normalized quantities), and every symbol with its meaning and nominal value.
  2. Build the equation map: which equation feeds which (e.g., model -> estimator -> score -> decision; geometry -> dimension -> rule). Note what each proposition explains.
  3. Note notation problems for the authors (same symbol used for two things, undefined symbols, inconsistent figure vs. text notation). Report them in a small "notation point" box.

Step 2 - Compute worked examples and verify

  • Pick one small, representative operating point from the paper (its nominal parameters). Write a Python script (numpy/scipy) that computes, for each equation, concrete example values (e.g., matrix entries, posterior mean/variance at a few indices, scores, probabilities, eigenvalues and cumulative energy, cost counts).
  • Look for a teaching example that shows WHY an equation is needed (e.g., two cases with the same mean but different outage; an adjacent next choice showing clustering).
  • Verify every closed-form result by Monte Carlo (about 1e6 samples) or direct computation, and record formula vs. simulation in a table.
  • Do NOT name a Python script numbers.py (it shadows the standard library module and breaks numpy). Use names like eq_numbers.py.

Step 3 - Helper drawings

With matplotlib (STIXGeneral font, mathtext stix, pdf.fonttype=42, about 6.6 x 3.3 in, 10-11 pt):

  • a plot that shows the meaning of a model equation (e.g., correlation vs. distance),
  • what the estimator believes after a few measurements (true value vs. predicted, and uncertainty),
  • how a probability formula depends on its inputs (with the teaching example marked),
  • spectrum / cumulative energy for a dimension definition. Add notes with ax.annotate(..., arrowprops=dict(arrowstyle='-|>', color='#C0392B'), bbox=dict(boxstyle='round,pad=0.3', fc='#FFF8E1', ec='#E09F3E')). Put notes in empty areas, increase ylim to create headroom, move legends away from data. Render each plot to PNG (pdftoppm -r 70) and check for overlaps before using it. For rank/size arguments, draw simple TikZ matrix shapes (wide vs. tall rectangle) instead of a plot.

Step 4 - Labeled equations (annotate-equations)

Use the LaTeX package annotate-equations (needs two or more pdflatex runs):

\usepackage{annotate-equations}
\renewcommand{\eqnhighlightheight}{\mathstrut}
\renewcommand{\eqnannotationfont}{\bfseries\small}
...
\vspace*{2.2em}
\begin{equation*}
\eqnmarkbox[teal]{mu}{\bm\mu}=\eqnmarkbox[blue]{c}{\mathbf C}\,\eqnmarkbox[orange]{a}{\mathbf A^{-1}}\,\eqnmarkbox[red]{y}{\mathbf y}
\end{equation*}
\annotate[yshift=1.2em]{above,left}{mu}{best guess}
\annotate[yshift=-1.2em]{below,right}{y}{measured data}
\vspace*{2.0em}

Rules that avoid collisions:

  • Wrap each labeled equation in a titled tcolorbox and reserve space with \vspace*{2.2em} before and \vspace*{2.0em} after (more if labels are stacked).
  • Alternate labels above/below and left/right; never put two labels on the same side at the same height next to each other. When needed, stack them with a larger yshift (e.g., 2.6em).
  • Keep labels short (3-6 words). Long equations: split into two displayed lines (e.g., mean on one line, covariance on the next), each with its own labels.
  • Labels must not cross the box edge: use left labels for left terms and right labels for right terms, and shorten text if needed.

Step 5 - Build the colored PDF

LaTeX article, A4, helvet sans, xcolor[table], tcolorbox[most], tikz (arrows.meta, positioning, calc, fit, backgrounds), booktabs, tabularx, xltabular, enumitem, titlesec, fancyhdr, float, caption, needspace, bm, annotate-equations. Colors: navy #1B3A5C, teal #00798C, rosered #D1495B, amber #E09F3E, leaf #2E8B57, plum #6A4C93. Boxes: verdict (teal), goodbox (green, small example), notebox (amber, where it comes from / notes), eqbox (navy title, the labeled equation). Define \newcolumntype{P}[1]{>{\raggedright\arraybackslash}p{#1}} and pass box titles as title={#1}.

Structure:

  1. Cover banner (navy band): "How to Read Every Equation in the Paper", paper title, date.
  2. Equation map (TikZ grid, absolute positions such as \def\dx{33mm}\def\dy{21mm}, boxes about 26 mm wide, straight arrows only, no crossings, short arrow labels) + one-line story + "how each section is built" note.
  3. All symbols in one table (symbol | simple meaning | value in paper) + notation-point box.
  4. One section per equation / proposition: labeled equation box -> "In plain words" -> "Where it comes from" (2-3 lines, when a derivation exists) -> helper drawing -> "Small example" with real numbers -> "Why the paper needs it".
  5. Cost/complexity formulas with an operation-count example.
  6. Table of the formulas behind the reported numbers (reduction %, CI with the t value, normalized metrics) with one example each from the paper.
  7. Checked by simulation table (formula vs. simulation) and a short statement of correctness; report any error found.
  8. Summary: one line per equation. Use \clearpage before major sections that hold big boxes and \Needspace{12\baselineskip} before smaller headings to avoid orphan headings.

Compile 3 times (pdflatex -interaction=nonstopmode). Missing packages: apt-get update && apt-get install -y texlive-latex-extra texlive-fonts-extra texlive-science texlive-pictures. If newtxmath fails with binhex.tex not found, use mathptmx.

Step 6 - Check before sending

  • Crop every equation box from the PDF (PyMuPDF page.search_for(title) then get_pixmap(clip=...)) and look at them stacked: no overlapping labels, no label outside the box, no text hidden.
  • Render all pages into a montage and check the equation map and page breaks.
  • Confirm all paper numbers match the paper and all example numbers match the script output.
  • Deliver the PDF with SendUserFile and summarize what each section contains, plus any notation issue or error found. Number new versions (v2, v3, ...) when the user asks for changes.

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.