latex-fmt
Skill182 starsChanged 42 days agoMITView on GitHub
Reformat LaTeX papers for a specified venue, year, track, and submission stage. Apply official templates and check layout, bibliography, page limits, and anonymization while preserving scientific content.
In this repository4 other files it gives its agents.
- latex-polishSkill
- latex-rescueSkill
- paper-readSkill
- pdf2texSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 8 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 24
Read the root document, build configuration, bibliography backend, and included files.
- Very specific: line 29
Read references/templates/venue-guide.md for official entry points and pitfalls.
- Distinctive: line 34
If authoritative rules are unavailable, preserve the existing format, identify unresolved requirements, and make any provisional …
- Very specific: line 36
Do not silently choose a previous year's kit or infer filenames by incrementing a year.
- Very specific: line 42
Follow the kit's example preamble, options, and author macros.
- Very specific: line 51
Do not invent neurips_2025.cls or icml2025.cls.
- Distinctive: line 56
Use the self-contained layout example when checking local widths and resolved references outside an official kit.
- Very specific: line 74
Check the exit status and final log, not just whether a PDF exists.
- Distinctive: line 80
Check the compiled main-content boundary against the verified limit.
- Distinctive: line 110
Use rules for the specified target rather than historical examples.
- Very specific: line 111
Keep official .cls, .sty, and .bst files intact.
- Distinctive: line 112
Do not remove author identities for camera-ready submissions by default.
- Distinctive: line 113
Never guarantee anonymity from a source keyword search alone.
- Distinctive: line 116
Do not submit, upload, or publish the paper as part of formatting.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Role
Apply the requested publication format using the matching official author kit and submission instructions. Distinguish successful compilation from verified submission compliance.
When to Activate
- The user asks to reformat a LaTeX manuscript or switch templates.
- The user asks for a pre-submission formatting check.
- The user invokes
/latex-fmtor$latex-fmt.
Workflow
Phase 1: Establish the target and rules
Read the root document, build configuration, bibliography backend, and included files. Establish the target venue, year, track/article type, and stage (anonymous review, preprint, or camera-ready). Ask only for missing information that changes the template or rules.
Read references/templates/venue-guide.md for official entry points and pitfalls.
Fetch the matching author kit and instructions, or use the user's supplied kit.
Record source URLs, check date, template filenames, page/word limits with
exclusions, anonymity rules, and required sections/checklists.
If authoritative rules are unavailable, preserve the existing format, identify
unresolved requirements, and make any provisional conversion's assumptions clear.
Do not silently choose a previous year's kit or infer filenames by incrementing
a year. This repository contains guidance, not official .cls, .sty, or .bst
files.
Phase 2: Apply the official template
Follow the kit's example preamble, options, and author macros. .cls files belong
in \documentclass; .sty files belong in \usepackage. For example, the
historical NeurIPS 2025 review pattern is:
\documentclass{article}
\usepackage{neurips_2025} % use the requested year's actual kitDo not invent neurips_2025.cls or icml2025.cls. Use the kit's bibliography
style rather than assuming the style package supplies a .bst. Remove custom
margin or spacing overrides only when they conflict with verified rules.
Consult formatting rules for layout guidance. Use the self-contained layout example when checking local widths and resolved references outside an official kit. It is an example, not a substitute for the requested venue's template. Preserve text, math, table values, citations, labels, and figure contents. Flag missing required sections for the author; do not generate unsupported ethical or empirical claims.
Phase 3: Bibliography and build
Respect the project's engine and bibliography backend. Prefer its existing build
command or latexmk when available. For manual builds, run the engine, the actual
BibTeX/Biber backend, and additional passes to resolve references. Do not run
pdflatex on a fontspec project or select Biber solely because biblatex is
present; check its backend option.
Preserve textual (\citet/\textcite) versus parenthetical (\citep/\parencite)
meaning when translating bibliography commands. Never blanket replace every
citation with \cite or \citep.
Check the exit status and final log, not just whether a PDF exists. Resolve
format-related errors with latex-rescue if available. Report missing tools or
template files as unverified compilation.
Phase 4: Length and anonymity
Check the compiled main-content boundary against the verified limit; total PDF pages cannot establish compliance when references or appendices are excluded. Count words according to the article type. Suggest cuts before moving sections or reducing figures. Avoid margin, font-size, or negative-spacing tricks.
For anonymous review, use the official review mode and inspect the main PDF, supplement, author blocks, acknowledgments, funding, self-citations, URLs, and PDF metadata. Flag identifying scientific text for contextual revision; keep citation keys intact. Remove identity metadata in source, rebuild, and inspect the result. An in-place metadata scrub does not prove anonymity. Camera-ready and preprint author information should follow their own rules.
Phase 5: Verification and report
Inspect the rendered PDF for clipped floats, readability, fonts, citations, and applicable anonymity requirements. Check checklist location and required sections against official rules. Report:
- Target venue/year/track/stage and authoritative sources with check date.
- Template and bibliography changes, with a reviewable diff.
- Build result, warnings, and actual page/word boundaries measured.
- Requirements as pass, fail, or unverified, with evidence.
- Missing author content or decisions and remaining actions.
A successful build is not a submission-ready claim. Use provisional/unverified status if a required rule, build, or visual check could not be confirmed.
Guardrails
- Preserve scientific content; do not invent citations or missing sections.
- Use rules for the specified target rather than historical examples.
- Keep official
.cls,.sty, and.bstfiles intact. - Do not remove author identities for camera-ready submissions by default.
- Never guarantee anonymity from a source keyword search alone.
- Preserve caption/author interfaces and textual versus parenthetical citations; package or command substitutions need verification in the actual kit.
- Do not submit, upload, or publish the paper as part of formatting.
Reference Files
references/templates/venue-guide.md— official venue entry points and template pitfalls.references/formatting-rules.md— layout, citation, and project organization guidance.
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.

