agentleFS
AlexyskoutnevAlexyskoutnev/research-skills/abstract-writing/SKILL.md

abstract-writing

Skill2 starsChanged 3 months agoMITView on GitHub

Write or revise a research-paper abstract using three master recipes distilled from dissecting top ML researchers' abstracts. Use when drafting an abstract, revising one for submission, or when the user asks how to structure or tighten an abstract. Applies the six-questions structure, the three recipes, and a mechanical checklist (short sentences, contribution first, no "can", no bare "This", numbers in pairs).

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 3 are distinctive, and 1 is 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.

  • 1

    Very specific

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

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

    Follow abstract-guide.md in this folder, it is the complete plain-English method.

    line 8
  2. Distinctive:

    Six answers first: Before any prose, write six rough bullets: what did you build / how does it work (ONE mechanism only) / where tested / …

    line 12
  3. Distinctive:

    Every "This/These" is followed by a noun ("This gate shows...", never "This shows...").

    line 25
  4. Distinctive:

    Every number has a partner (from X to Y.

    line 26

≈ “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.

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

Abstract Writing (three master recipes)

Follow abstract-guide.md in this folder, it is the complete plain-English method. This file is the operating summary.

Process

  1. Six answers first. Before any prose, write six rough bullets: what did you build / how does it work (ONE mechanism only) / where tested / numbers (as pairs) / the control-ablation / why it matters. One or two sentences each is the whole abstract.
  2. Pick a recipe from the guide: A (short + confident: one crisp mechanism, no numbers, comparative claims against baseline classes), B (numbers as drama: tension opener, paired before/after figures, kill the reviewer's doubt in one clause), C (anchor then extend: start from a familiar idea, enumerate concretely, end on an idea not a metric). Default for results-heavy systems papers: B's structure with A's compression. If the advisor favors ideas over metrics, close with C's idea sentence.
  3. Draft, then run the mechanical audit (never skip):
    • Sentence 1 states the contribution with a strong verb (give, present, introduce).
    • Count words per sentence: none above ~28, average ~20.
    • Zero occurrences of "can", "cannot", "could".
    • Every "This/These" is followed by a noun ("This gate shows...", never "This shows...").
    • Every number has a partner (from X to Y; ours vs. theirs; with vs. without).
    • Each number is attached to the correct claim, never borrow a similar-looking statistic from a different measurement.
    • Last sentence is an idea, not a number. 140-190 words. No contractions.
  4. Read it aloud; fix every stumble. Compare against the closest rival paper's abstract and make the differentiation audible in one clause.

Reference file

  • abstract-guide.md: the full plain-English guide (recipes, templates, sentence surgery, the assembled fill-in-the-blank template, and twelve further rules from the classic writing guides and canonical ML abstracts).

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.