sci-abstract
GitHub专为Science期刊设计的摘要与一句话总结润色技能。针对方法回顾型或超字数稿件，提供符合AAAS规范的量化结果导向写作指导，确保内容对非专业科学家可读，提升编辑初审通过率。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill sci-abstract -g -y
SKILL.md
Frontmatter
{
"name": "sci-abstract",
"description": "Use when writing the two front-matter artifacts Science (AAAS) demands — the one-sentence summary for the table of contents and the ≤125-word single-paragraph abstract — both readable by any scientist and both leading with the quantified advance rather than the method. Late-stage polish skill for the flagship AAAS weekly."
}
Abstract & One-Sentence Summary (sci-abstract)
When to trigger
- Significance, framing, and format are settled (do this late).
- The abstract reads like a method recap with no result.
- There is no one-sentence summary, or it restates the title.
- The abstract exceeds ~125 words or is dense with jargon/acronyms.
Two distinct artifacts Science requires
1. One-sentence summary (≤ ~125 characters)
A single declarative sentence for the table of contents and editors. It is not the title and not the first line of the abstract.
- States the finding and its general point.
- No undefined acronyms; readable by any scientist.
- Active voice. Example shape: "A single mutation rewires metabolism, explaining cold tolerance across insects."
2. Abstract (≈ 100–125 words, unstructured prose)
Science abstracts are short, single-paragraph, no subheadings. Target the educated general scientist, not the subfield.
Recommended five-move structure (no labels in the text):
- Context / stakes (1 sentence) — the broad problem.
- Gap / question (1 sentence) — what was unknown.
- What we did (1 sentence) — approach, in plain terms.
- Key result, quantified (1–2 sentences) — the advance, with numbers.
- Implication (1 sentence) — why a broad community should care.
Hard constraints
- ≤ ~125 words (check the current author guidelines for the exact cap; treat 125 as the ceiling).
- No subheadings, no citations, no figure/table references in the abstract.
- Define any acronym on first use, or avoid it.
- At least one quantified result — not "significantly increased" but "increased 2.4-fold (P < 0.001)".
- First sentence is comprehensible to a scientist outside the field.
Jargon blacklist (rewrite on sight)
- "Herein we report…", "Importantly,", "Interestingly,", "Notably,"
- Strings of ≥2 undefined acronyms in one sentence.
- "elucidate", "delineate", "interrogate", "leverage" as filler verbs.
- Hedging stacks: "may potentially suggest that it could…".
Quantification check
Every claim of effect must carry a number somewhere in the paper, and the headline effect should be in the abstract: magnitude + unit + uncertainty (CI or P). If the abstract has zero numbers, it is not finished.
What the editorial board reads first
Because Science triages most submissions in-house before any referee sees them, your one-sentence summary and abstract are frequently the only prose a professional editor reads before deciding to read further. Calibrate to that reality:
| Editor's silent question | Where they look | What kills it |
|---|---|---|
| "Would our general readership care?" | one-sentence summary | a subfield claim only specialists parse |
| "Is there a result, or just an approach?" | abstract sentences 3–4 | method recap with no quantified finding |
| "Is this overstated?" | abstract last sentence | an implication asserted, not demonstrated |
| "Can I send this summary to a referee verbatim?" | the whole summary | undefined acronyms, hedging stacks |
The abstract is one unstructured paragraph — unlike a structured clinical abstract (Background/Methods/Results/Conclusions). Do not import a NEJM/JAMA structured-abstract template; Science runs continuous prose with no labels.
Worked micro-example (illustrative)
A team finds that a single gut bacterium's enzyme degrades a common drug, explaining why some patients respond and others do not. Walk the moves (numbers illustrative only):
- One-sentence summary (≈110 chars): "A single gut-microbial enzyme inactivates a widely used drug, explaining variable patient response."
- Abstract (≈120 words, prose): Stake — drug response varies unpredictably between patients. Gap — the gut microbiome was suspected but no mechanism was known. Approach — we screened 412 isolates and reconstituted the activity in vitro. Result — one enzyme reduced active-drug levels by 73% (95% CI 65–80%; P < 0.001), and its abundance predicted response in a 96-patient cohort (AUC 0.84). Implication — microbial enzymology may be a general, modifiable axis of drug variability.
Notice: the headline number (73%, CI, P) sits in the abstract, not buried in a figure; the last sentence reaches a broad audience (pharmacology, microbiology, clinical medicine) without claiming proof of causation in humans.
Referee and editor pushback patterns
| Pushback you will hear | The Science-specific fix |
|---|---|
| "The summary just restates the title." | Make it state the consequence, not the topic — the lesson a non-specialist takes away. |
| "I read the abstract and still don't know the result." | Put the quantified headline (magnitude + uncertainty) in sentences 4–5; delete one context sentence to fit. |
| "Reads like an abstract for a specialist journal." | Replace the field-internal opener with the broad stake; cut acronym chains. |
| "The last line over-claims." | Hedge to what the data show; "may be a general axis" not "is the cause" — confirm the claim survives sci-fit. |
Calibration anchors (confirm against current author guidelines)
- One-sentence summary: target ≤ ~125 characters; it is a distinct required field, not the first abstract line.
- Abstract: target ≤ ~125 words, single paragraph, no subheadings, no citations, no figure references.
- Exactly one paragraph — the convention is firm even as the word ceiling may shift; verify the current cap.
- At least one quantified headline effect with uncertainty; zero-number abstracts are unfinished.
Output format
【One-sentence summary】 "..." (char count: N ≤ 125)
【Abstract】 single paragraph (word count: N ≤ 125)
【Five moves present?】 context / gap / approach / quantified result / implication
【Quantified headline result?】 yes/no + the number
【Jargon hits removed】 [...]
【Next】 sci-citation
Anti-patterns
- Do not make the one-sentence summary a paraphrase of the title.
- Do not open the abstract with method or organism; open with the stake.
- Do not end on "further work is needed" — end on the implication.
- Do not pad to sound comprehensive; the cap is a feature.
Version History
- 1839142 Current 2026-07-05 14:24
