polish-tables-figures
Skill59 starsChanged 5 months agoApache-2.0View on GitHub
Polishes LaTeX tables and figures to publication quality. Use when a researcher mentions table formatting, booktabs, figure layout, subfigures, captions, colorblind-safe colors or palettes, cref/cleveref consistency, overfull hbox, column widths, or asks to make tables and figures look professional or camera-ready for a venue (ACM, IEEE, NeurIPS, LNCS, CHI). Converts hline-and-vertical-rule tables to booktabs, sizes columns within the page budget, lays out subfigures with subcaption, applies caption conventions per venue family (table captions above, figure captions below, ACM Description alt-text), fixes ref/cref inconsistencies, and recommends colorblind-safe palettes (Okabe-Ito, Paul Tol) using bundled stdlib-Python linters and palette generators. Edits presentation only — never alters data values, units, or results.
In this repository43 other files it gives its agents.
- docs/llms-full.txtllms.txt
- docs/llms.txtllms.txt
- add-venue-profileSkill
- anonymize-paperSkill
- assess-paperSkill
- benchmark-paperSkill
- check-originalitySkill
- draft-related-workSkill
- draft-surveySkill
- fetch-paperSkill
- find-papersSkill
- fit-page-limitSkill
- literature-reviewSkill
- make-posterSkill
- make-slidesSkill
- match-styleSkill
- orchestrate-paperSkill
- paper-profileSkill
- parse-cfpSkill
- plan-submissionSkill
- polish-proseSkill
- preflight-checkSkill
- prepare-artifactsSkill
- prepare-camera-readySkill
- refactor-research-codeSkill
- refactor-structureSkill
- reflect-and-improveSkill
- reflect-paperSkill
- rehearse-qaSkill
- render-workspace-htmlSkill
- select-venueSkill
- simulate-reviewersSkill
- study-exemplarsSkill
- tailor-to-venueSkill
- test-research-codeSkill
- triage-reviewsSkill
- verify-citationsSkill
- verify-claimsSkill
- verify-resultsSkill
- work-with-overleafSkill
- write-abstractSkill
- write-rebuttalSkill
- write-talk-scriptSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 3 are distinctive, and 2 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.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 21
After tailor-to-venue, before preflight-check.
- Very specific: line 35
Resolve venue conventions: If a venue is named, load its profile (sanity-check with python3 scripts/venueyaml.py …
- Distinctive: line 79
Check the colors actually used in the paper (hex values from TikZ/pgfplots definitions, or sampled from included plots):
- Distinctive: line 118
Never fabricate venue rules.
- Distinctive: line 120
Do not regenerate or recolor result plots yourself.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Polish Tables & Figures
Turn draft-quality LaTeX floats into the tables and figures reviewers expect
to see at ACM / IEEE / NeurIPS-style / LNCS venues: booktabs rules, columns
that fit the page budget, clean subfigure layouts, venue-correct captions,
consistent \cref references, and colorblind-safe colors. Every change is
presentation-only; the numbers are sacred.
When to use
- "Make my tables/figures look professional" / "polish for camera-ready"
- "Convert this table to booktabs" / "my table is too wide" / "overfull hbox"
- "Lay these plots out as subfigures" / "fix my captions"
- "Are my figure colors colorblind-safe?" / "pick a palette for this plot"
- "Fix my \ref/\cref mess"
- After
tailor-to-venue, beforepreflight-check.
Inputs
- The main
.texfile (the one with\documentclass);\input/\includefiles are followed automatically. - Optional: a venue profile
venues/conferences/<venue>-<year>.yml(schema invenues/schema.yml) — sets column count and caption conventions. Without one, conventions are inferred from the documentclass. - Strongly recommended: a compiled
.lognext to the.texso the column-sizing (overfull box) check can run.
Process
- Resolve venue conventions. If a venue is named, load its profile
(sanity-check with
python3 scripts/venueyaml.py venues/conferences/<venue>.yml). Re-verify format-critical facts against the livecfp_urlbefore relying on them: template/documentclass, column count, any stated figure/table rules (minimum font sizes, color/grayscale requirements, accessibility requirements). Profiles are a starting point, never ground truth. Lint everything first:
python3 scripts/check_floats.py paper.tex \ --venue venues/conferences/<venue>-<year>.ymlFlags:
--json,--strict(warnings also fail),--log <file>,--overfull-threshold <pt>,--no-inputs. Exit codes: 0 clean, 1 findings, 2 bad arguments. The linter covers: booktabs style (\hline,\cline, vertical rules, double rules, missing package),\resizebox/tiny-font scaling, caption position per family, missing\caption/\label/\centering/ACM\Description, label-before-caption, deprecatedsubfigurepackage,width=\textwidthin a one-column float at a two-column venue, raster graphics,[h]/[H]placement, undefined and unreferenced labels, mixedFigure~\refvs\crefstyles, "Figure \cref" double prefixes, cleveref load order, lowercase\crefat sentence start, and Overfull\hboxentries from the.log.- Fix tables — work through findings using
references/tables.md: the mechanical
booktabs conversion recipe (it preserves cell content byte-for-byte),
then column sizing within the page budget (cut → abbreviate → align
numbers →
\tabcolsep→tabularx→table*— never\resizebox). - Fix figures — references/figures.md:
width discipline (
\columnwidth/\linewidth,figure*for spans), subfigure layout withsubcaption, vector formats, font sizes inside plots, float placement. - Fix captions and cross-references —
references/captions-crossrefs.md:
per-family caption position/style table, self-contained caption writing,
and the cleveref setup (load last,
capitalise,\crefvs\Cref, one style everywhere). Fix color — references/color-accessibility.md. Check the colors actually used in the paper (hex values from TikZ/pgfplots definitions, or sampled from included plots):
python3 scripts/palettes.py check 4477AA EE6677 ... # arbitrary colors python3 scripts/palettes.py show okabe-ito --format latex|pgfplots|matplotlibchecksimulates protanopia/deuteranopia/tritanopia and grayscale print, flags confusable pairs (exit 1), and the references file explains redundant encoding (markers + line styles, not color alone). If plots are generated by user scripts (matplotlib/pgfplots), edit the plotting source and ask the user to regenerate — never claim recolored results.- Recompile and re-run the linter until the verdict is PASS (or every
remaining WARN is a deliberate, user-approved choice). Then hand off to
preflight-checkfor the full desk-reject lint.
Output
Edited .tex (and plotting-script) changes applied in place with a short
change log per float: what changed and why, citing the venue convention.
Plus the final linter report (PASS / PASS-WITH-WARNINGS / FAIL). No new
files unless the user asks for extracted style snippets (e.g. a
colors.tex palette preamble).
Adapt to your discipline
Conventions here target CS venues (ACM/IEEE/ML/LNCS). Forking for another
field: adjust the caption-position conventions in
references/captions-crossrefs.md and the family mapping in
scripts/check_floats.py (convention_label); the booktabs, sizing, and
color guidance is field-agnostic.
Guardrails
- Never change data: cell values, units, significant digits, axis scales, error bars, and legends' meanings are untouchable. Reformatting only. If a value looks wrong, tell the user — do not "fix" it.
- Never fabricate venue rules. The profile and the live CFP are the only sources; when neither states a rule, label advice as general convention.
- Do not regenerate or recolor result plots yourself; edit the user's plotting source and let them re-run it.
- Bold-best-result formatting must come from the user's stated criterion, not your reading of which number "looks best".
- Citation issues found along the way go to
verify-citations; submission readiness goes topreflight-check. Never submit anything on the user's behalf.
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.

