agentleFS
Sign inStart free
acprkacprk/crypto-research-agents/skills/abstract-craft/SKILL.md

abstract-craft

Skill15 starsChanged 9 days agoMITView on GitHub

Use when writing or tightening a paper abstract (and its plain-text submission-form twin). Applies the three-part paradigm (state of the art, our technique, results and artifact) with a sentence-level function table and an overclaim filter.

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:

    every number → EVIDENCE row id noted in a LaTeX comment % EV:E12.

    line 25
  2. Distinctive:

    every scope noun ("13 rings", "all primes p ≡ 1 mod 4") → matches the parameter/security tables.

    line 26
  3. Distinctive:

    Never mix instances below the claimed security level into the range.

    line 61
  4. Distinctive:

    If a baseline number is reproduced rather than taken from the paper, the abstract compares against the reproduced number and the paper …

    line 62

≈ “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 craft

Inputs

CLAIMS.md (survived/weakened claims), EVIDENCE.md (numbers), the contribution list (paper/sections/intro.tex), target venue (for length norms, see venue-calibration).

Procedure

  1. Write a one-line viewpoint of the paper: "We view X as Y." If this sentence cannot be written, the abstract is not ready; go back to the overview.
  2. Fill the sentence-function table below in paper/ABSTRACT_PLAN.md (Chinese or English bullet map is fine; discuss it with the human author before prose).
  3. Draft three paragraphs (see paper-playbook/sections/02_abstract.md).
  4. Overclaim filter (automatic pass, then manual):
    • forbidden: novel, groundbreaking, rigorous guarantee, optimal (unless proven), significantly (without a number), first (unless a literature search in LITERATURE.md supports it, and even then prefer "to our knowledge, the first").
    • every number → EVIDENCE row id noted in a LaTeX comment % EV:E12.
    • every scope noun ("13 rings", "all primes p ≡ 1 mod 4") → matches the parameter/security tables.
  5. Consistency pass: abstract ↔ contributions ↔ conclusion use the same numbers, the same order and the same framework name.
  6. Length: IACR LNCS: fits page 1 with keywords (typically 200–300 words). ToSC/TCHES: usually 200–280 words. Security conferences: often ≤ 200–250 words in the submission form. Check the CFP.
  7. Plain-text twin paper/abstract_plain.txt for the submission form.

Sentence-function table (7–10 sentences)

#functionnotes
S1field consensus + bottleneckno citation needed
S2existing routes, grouped, venue'year1–2 representatives per route
S3open question"However, ... has remained open."
S4our object and viewpoint"We present/propose X, which views ... as ..."
S5–S6technical points, one each"First, ... At the same time, ..." or structure → bound → construction
S7what we prove (property, not activity)not "we analyse security" but "we prove that no attacker learns ... from three or fewer outputs"
S8scope → range → concrete points vs named baselinemeasured on the same machine, if true, say so
S9artifact / formalisation / invitation for analysisexact scope of formal verification

Upgrading weak sentences

weakstrong
"We give a complete security analysis.""Every considered attack needs more than 2^128 operations; the tightest margin comes from <attack> at 2^{x}."
"Our method is much faster.""On N instances at ≥λ-bit security, <stage> is 2.1–3.4× faster than <baseline> in the same library."
"We formally verify our results in Lean.""The algebraic core (Lemmas 2–5) is machine-checked in Lean 4; the noise analysis is not."
"We unify prior work.""The methods of [A] and [B] are the cases w=1 and k=1 of our construction."

Headline-range hygiene

  • Lower end = the most conservative measured number; if the upper end is an estimate, label it as an estimate in the table and say "up to ... (estimated)".
  • Never mix instances below the claimed security level into the range.
  • If a baseline number is reproduced rather than taken from the paper, the abstract compares against the reproduced number and the paper reports both.

Output

paper/sections/abstract.tex, paper/abstract_plain.txt, and a short diff note in paper/CHANGELOG.md.

See toy examples: skills/paper-playbook/examples/toy_abstracts.md.

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.

Check your setup, free