Top cs polishing

Skill zhiming33416/top-cs-paper-skills/skills/top-cs-polishing

Evidence-grounded Codex skills for writing, polishing, reviewing, reviewer responses, and figures for top CS conferences.

Install

npx -y skills add zhiming33416/top-cs-paper-skills --skill top-cs-polishing

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

2 things to look at

24 days oldThe repository was created 24 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
1 stars1 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

Polish, translate, tighten, structurally revise, or diagnose existing computer-science manuscript prose and LaTeX for WWW, ICLR, ICML, or a generic venue while preserving evidence. Use for Chinese-to-English academic rewriting, paragraph flow, section-level revision, claim calibration, concision, terminology consistency, reducing generic AI prose, revision ledgers, and LaTeX layout or float-placement problems. Do not use to invent a paper from sparse notes, simulate peer review, or write a reviewer response.

SKILL.md

2.8 KB, as published. Nobody here has run it

Top CS Paper Polishing

Route the request

Read manifest.yaml and every always_load path.
Detect and state venue, paper type, section, language, revision mode, artifact scope, evidence state, submission stage, citation verification, and figure handoff.
Load only the static fragments mapped to those values.
Open deeper references only when an on_demand condition applies.
Use generic for unsupported venue/year combinations and warn that official rules require verification.

Execute

Extract immutable facts: quantities, comparison direction, conditions, citations, equations, uncertainty, and explicit boundaries.
Build or update the terminology ledger for multi-paragraph work.
Diagnose in this order: framing, rhetorical structure, evidence connection, terminology, and sentence form.
Respect the selected revision mode: keep light-edit local, permit reordering in structural-revision, and prove information preservation in compression.
Revise in the order paper role -> section job -> paragraph logic -> claim/evidence/boundary -> sentence form.
Record every material change and any claim-strength, citation, or figure-reference shift in the revision ledger.
Verify preserved facts, citations, LaTeX, terminology, and reader workflow; report what was compiled, rendered, structurally checked, or not verified.

Preserve scientific meaning and never silently strengthen novelty, causality, generality, or significance. Preserve LaTeX commands, labels, citations, equations, and comments unless a necessary change is explicitly reported. For a supplied LaTeX project, run scripts/check_latex_project.py when an engine is available; never enable shell escape or write build artifacts into the source project.

When auditing figures, check claim, caption, callout, terminology, and label alignment only. Return a corrected handoff brief when necessary; do not render or restyle the figure. Metadata verification does not establish that a cited source supports the manuscript claim.

For page-layout or float-placement requests, load references/latex-layout.md, compile to an isolated output directory, and inspect rendered pages before and after. Do not judge layout from source alone.