30 Claude Prompts for Abstracts You Can Defend

<context> I am submitting a conference talk abstract to [CONFERENCE]. I will paste a draft or honest notes. Do not invent completed results I have not run. </context> <inputs> - Draft or notes: [PASTE NOTES] - Conference: [CONFERENCE] - Word limit: [WORD LIMIT] - Work status: [COMPLETED / IN PROGRESS / PLANNED] - Session type: [TALK / POSTER / UNKNOWN] - Title I am considering: [WORKING TITLE OR NONE] </inputs> <task> Draft a talk abstract within [WORD LIMIT]: 3 title options, the abstract, and a one-line status of the work. If results are not in the paste, do not write them as done. Prefer present or future verbs when status is IN PROGRESS or PLANNED. </task> <constraints> - Do not invent papers, data, field sites, n, or citations. - Do not name coauthors I did not list. - Do not invent a session, track, or chair. - If a claim needs a source I did not give, mark it UNVERIFIED. - Count words and stop at the limit. </constraints> <format> Three titles, abstract, status line, word count, UNVERIFIED rows. </format>

<context> I need a poster abstract that a walker can parse. It must still be honest. Do not invent a figure, a sample, or a finding to make the poster look finished. </context> <inputs> - Notes or draft: [PASTE NOTES] - Conference: [CONFERENCE] - Word limit: [WORD LIMIT] - Work status: [COMPLETED / IN PROGRESS / PLANNED] - What the poster will show if I know: [FIGURES I HAVE OR NONE] </inputs> <task> Write a poster abstract within [WORD LIMIT]: problem, method I actually used or will use, findings only if pasted, and one takeaway I can defend. If [FIGURES I HAVE OR NONE] is NONE, do not promise a chart. Mark missing cells as NOT IN PASTE. </task> <constraints> - Do not invent data, n, a figure, or a citation. - Do not write past-tense results for IN PROGRESS or PLANNED work. - Do not upgrade a planned analysis into a completed table. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Poster abstract, word count, NOT IN PASTE list, UNVERIFIED rows. </format>

<context> The study is not finished. I need a conference abstract that tells the truth about status. Do not invent results so the abstract reads like a completed paper. </context> <inputs> - Notes on aims, methods, and what is done: [PASTE NOTES] - Conference: [CONFERENCE] - Word limit: [WORD LIMIT] - Work status: [IN PROGRESS / PLANNED] - What I will have by the meeting: [EXPECTED BY MEETING OR UNKNOWN] </inputs> <task> Draft an in-progress abstract within [WORD LIMIT]: question, design I can run, work completed so far from the paste, and what remains. Use future or present-in-progress verbs. If [EXPECTED BY MEETING OR UNKNOWN] is UNKNOWN, do not promise findings by the meeting date. </task> <constraints> - Do not invent papers, data, n, or completed results. - Do not write "we found" for work I have not run. - Do not invent a meeting date or a data-release date. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Abstract, verb-tense audit, will-not-claim list, UNVERIFIED rows. </format>

<context> I need a conference abstract that states why this work fits a session I can name. Do not invent a track, chair, or theme I did not paste. </context> <inputs> - Notes or draft: [PASTE NOTES] - Conference: [CONFERENCE] - Session or track I am targeting: [SESSION TYPE] - Why I think it fits, in my words: [FIT REASONS] - Word limit: [WORD LIMIT] </inputs> <task> Write the abstract within [WORD LIMIT] from [PASTE NOTES]. Add one or two sentences of session fit using only [FIT REASONS] and [SESSION TYPE]. If my reasons are thinner than a fit claim, shrink the fit sentence or mark it UNVERIFIED. </task> <constraints> - Do not invent a track name, chair, or call-for-papers line I did not paste. - Do not invent papers that appeared in last year's session. - Do not invent data to look more on-theme. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Abstract, fit sentence with a defend or shrink tag, UNVERIFIED rows. </format>

<context> The conference wants a short abstract and a longer one. Both must say the same study. Do not add findings in the long version that the short version cannot support. </context> <inputs> - Notes or draft: [PASTE NOTES] - Conference: [CONFERENCE] - Short word limit: [SHORT LIMIT] - Long word limit: [LONG LIMIT] - Work status: [COMPLETED / IN PROGRESS / PLANNED] </inputs> <task> Write two abstracts from the same paste: one at [SHORT LIMIT] and one at [LONG LIMIT]. The long version may add method detail and a limit that are already in the paste. It may not add a result, n, or citation that the short version lacks unless that fact is in [PASTE NOTES] and you flag it as DETAIL FROM PASTE. </task> <constraints> - Do not invent papers, data, or results in either version. - Do not let the long abstract become more certain than the short one. - Honor [COMPLETED / IN PROGRESS / PLANNED] in both. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Short abstract, long abstract, both word counts, DETAIL FROM PASTE list, UNVERIFIED rows. </format>

<context> I need a structured IMRaD abstract from notes I have. Fill only the headings I list. Do not invent a finding to complete a heading. </context> <inputs> - Notes or draft: [PASTE NOTES] - Headings I must use: [ABSTRACT HEADINGS] - Word limit: [WORD LIMIT] - Venue: [VENUE] - Sample n if known: [SAMPLE N OR UNKNOWN] </inputs> <task> Write a structured abstract within [WORD LIMIT] under [ABSTRACT HEADINGS] only. Every sentence must be traceable to [PASTE NOTES]. If a heading has no support, write NOT IN DRAFT under it. Leave n as UNKNOWN unless I pasted it. </task> <constraints> - Do not invent papers, authors, years, data, n, p-values, or implications. - Do not add a heading I did not list. - Do not make Results more positive than the notes. - If a claim needs a source I did not give, mark it UNVERIFIED. - Count words across all headings and stop at the limit. </constraints> <format> Structured abstract under my headings, NOT IN DRAFT list, word count, UNVERIFIED rows. </format>

<context> I need Background and Objective lines for a structured abstract. Do not open with an invented statistic or a famous paper I did not paste. </context> <inputs> - Research question: [RESEARCH QUESTION] - Notes I have: [PASTE NOTES] - Gap I can defend: [GAP I CAN DEFEND] - Venue: [VENUE] - Word budget for these headings: [WORD LIMIT] </inputs> <task> Draft Background and Objective only. Background states the problem from my notes. Objective states [RESEARCH QUESTION] in one or two sentences. If [GAP I CAN DEFEND] is stronger than the notes, weaken the gap sentence. Do not survey a literature I have not pasted. </task> <constraints> - Do not invent papers, authors, years, prevalence numbers, or quotations. - Do not write "scholars have long debated" unless my paste supports a named debate. - If a sentence needs a source I did not paste, mark it UNVERIFIED or drop it. - Stay inside [WORD LIMIT] for these two headings together. </constraints> <format> Background, Objective, claim-to-source checklist, UNVERIFIED rows. </format>

<context> I will paste the protocol I can describe. The Methods heading must match that paste. Do not complete a missing n from a typical paper in my field. </context> <inputs> - Protocol or methods notes: [PASTE PROTOCOL] - Sample n if known: [SAMPLE N OR UNKNOWN] - Design: [STUDY DESIGN] - Venue: [VENUE] - Word limit for Methods: [WORD LIMIT] </inputs> <task> Write the Methods heading from [PASTE PROTOCOL]: design, participants or corpus, collection, and analysis as stated. If [SAMPLE N OR UNKNOWN] is UNKNOWN, keep every sentence n-free and list the cells I must fill. Use past tense only for work I marked as done in the paste. </task> <constraints> - Do not invent n, a response rate, an instrument, or a second coder. - If a cell is missing, write NOT IN PASTE. - Do not upgrade a convenience sample into a representative one. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Methods heading, known vs NOT IN PASTE table, UNVERIFIED rows. </format>

<context> I need a Results heading written from findings I already have. Do not invent a percent, an n, or a p-value to make the heading look complete. </context> <inputs> - Findings I can defend: [PASTE FINDINGS] - Sample n if known: [SAMPLE N OR UNKNOWN] - Venue: [VENUE] - Word limit for Results: [WORD LIMIT] - Hedges I already use: [HEDGES I ALREADY USE] </inputs> <task> Write the Results heading using only numbers and claims in [PASTE FINDINGS]. Keep my hedges. If a comparison or subgroup is not in the paste, omit it. If [SAMPLE N OR UNKNOWN] is UNKNOWN, do not insert an n. </task> <constraints> - Do not invent n, percents, effect sizes, p-values, or quotations. - Do not upgrade "suggests" to "shows" for polish. - Do not add a finding I mentioned in conversation but did not paste. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Results heading, number-by-number source map, omitted comparisons, UNVERIFIED rows. </format>

<context> The Conclusion heading must match the results I pasted, not a bigger paper. No new case, no new number, no stronger verb. </context> <inputs> - Results I can defend: [PASTE RESULTS] - Limits I already accept: [PASTE LIMITS] - Objective as I stated it: [RESEARCH QUESTION] - Venue: [VENUE] - Word limit for Conclusion: [WORD LIMIT] </inputs> <task> Write a Conclusion of about [WORD LIMIT] that restates what [PASTE RESULTS] can support, names a limit from [PASTE LIMITS], and does not outrun [RESEARCH QUESTION]. If the results are thinner than the objective, shrink the conclusion. Flag any sentence that still needs a source as UNVERIFIED. </task> <constraints> - Do not invent papers, data, policy wins, or a next study with a fake design. - Do not introduce a finding that is not in [PASTE RESULTS]. - Prefer "this set cannot speak to X" over "the field has never." - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Conclusion heading, mismatch list (objective vs results), UNVERIFIED rows. </format>

<context> I need a lay summary for a funder that stays faithful to an abstract or notes I pasted. Clarity is not a license to overclaim or to invent an impact statistic. </context> <inputs> - Source text: [PASTE ABSTRACT] - Funder: [FUNDER] - Word limit: [WORD LIMIT] - Who is affected, from my notes: [AFFECTED GROUP OR UNKNOWN] - Required headings if any: [FUNDER HEADINGS OR NONE] </inputs> <task> Rewrite the core claim, method as stated, and conclusion as stated in short sentences a non-specialist reviewer can follow. Stay inside [WORD LIMIT]. If [AFFECTED GROUP OR UNKNOWN] is UNKNOWN, do not invent a population. Mark any impact line that still needs a source as UNVERIFIED. </task> <constraints> - Do not invent papers, prevalence numbers, jobs saved, or policy wins. - Do not upgrade "suggests" to "proves" for readability. - Do not invent a human-interest anecdote. - If a number is not in the paste, omit it. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Lay summary, word count, terms you simplified with the original wording, UNVERIFIED rows. </format>

<context> I need a public or patient lay abstract that a [READER] can follow. Do not invent an anecdote, a risk, or a treatment claim the paste does not make. </context> <inputs> - Source text: [PASTE ABSTRACT] - Reader: [READER] - Word limit: [WORD LIMIT] - Terms I must keep: [TERM LIST OR NONE] - What I must not imply: [OFF LIMITS] </inputs> <task> Rewrite [PASTE ABSTRACT] for [READER] in about [WORD LIMIT]: what was asked, what was done as stated, what was found as stated, and what it does not mean. Define one or two terms if the paste uses them. Honor [OFF LIMITS]. Keep [TERM LIST OR NONE] when I listed terms. </task> <constraints> - Do not invent a patient story, a clinic name, or a dosage. - Do not add a famous paper as background. - Do not turn a lab finding into a care recommendation I did not paste. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Lay abstract, glossary of simplified terms, OFF LIMITS check, UNVERIFIED rows. </format>

<context> The journal wants a plain-language summary beside the scientific abstract. It must say the same study. Do not invent a rule or a finding to satisfy the box. </context> <inputs> - Scientific abstract or draft: [PASTE ABSTRACT] - Venue rules I copied: [VENUE RULES] - Word limit: [WORD LIMIT] - Reader the journal names: [READER OR PUBLIC] </inputs> <task> Draft a plain-language summary within [WORD LIMIT] that follows [VENUE RULES] I pasted. Same claims, same hedges, shorter words. If a rule in the paste needs a fact I do not have, write NOT IN PASTE instead of filling it. </task> <constraints> - Do not invent a journal rule I did not paste. - Do not invent data, citations, or a wider impact. - Do not upgrade claim strength for readability. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Plain-language summary, word count, rules you could not meet, UNVERIFIED rows. </format>

<context> I need a one-paragraph impact statement from notes I can defend. Do not open with an invented statistic, citation count, or policy fact. </context> <inputs> - Notes on who is affected and why: [PASTE NOTES] - Research question: [RESEARCH QUESTION] - Word limit: [WORD LIMIT] - Audience: [FUNDER / JOURNAL / DEPARTMENT] - Citation style if cites are allowed: [CITATION STYLE OR NONE] </inputs> <task> Write one impact paragraph within [WORD LIMIT]: who is affected only if the paste says so, what is still open in the notes, and what this project can touch. Flag any sentence that still needs a source I must find as UNVERIFIED. </task> <constraints> - Do not invent papers, authors, years, prevalence numbers, or citation counts. - If a why-it-matters claim has no support in [PASTE NOTES], rewrite it as a question or drop it. - Do not promise an impact my design cannot produce. - If [CITATION STYLE OR NONE] is NONE, use zero citations. </constraints> <format> Impact paragraph, claim-to-source checklist, UNVERIFIED rows. </format>

<context> I need a classroom or outreach summary of work I already wrote. This is not a new argument and not a press release with extra wins. </context> <inputs> - Source text: [PASTE ABSTRACT] - Reader: [READER] - Word count: [WORD COUNT] - Setting: [CLASS / MUSEUM / NEWSLETTER / OTHER] - Terms I must keep: [TERM LIST OR NONE] </inputs> <task> Summarize [PASTE ABSTRACT] in about [WORD COUNT] for [READER] in [CLASS / MUSEUM / NEWSLETTER / OTHER]. Keep the actual claims and hedges. Define terms only from the paste. Do not add a source, a finding, or a slogan that is not in the paste. </task> <constraints> - Do not invent an anecdote, a celebrity, or a local hook. - Do not strengthen hedges into proofs. - Do not add citations. - If the source is confused, say so in a note after the summary. Do not clean it into a better study than I wrote. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Outreach summary, word count, terms simplified, UNVERIFIED rows. </format>

<context> I need titles drawn from an abstract or draft I actually have. They must name the paper I wrote, not a trendier paper. </context> <inputs> - Abstract or draft: [PASTE ABSTRACT] - Working thesis or question: [WORKING THESIS] - Venue limits: [VENUE LIMITS] - Words I must include if any: [MUST INCLUDE OR NONE] </inputs> <task> Propose 5 titles. Every content word should appear in the paste or be a plain synonym of a paste term. Rank titles by accuracy to [WORKING THESIS], not by punch. Honor [VENUE LIMITS]. Flag any title that would imply a method, population, or theory I did not write. </task> <constraints> - Do not invent a method, population, dataset, or theory name that is not in the paste. - Do not stuff a colon subtitle with a fashionable framework I did not use. - If the paste is thin, give fewer titles and say why. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Ranked titles with a one-line accuracy note, rejected trendy words, UNVERIFIED rows. </format>

<context> I need a keyword list drawn from text I wrote. Keywords must name this paper. Do not add a journal, author, or topic I did not study. </context> <inputs> - Abstract or draft: [PASTE ABSTRACT] - Venue limits: [VENUE LIMITS] - Terms I already use: [TERM LIST] - Controlled vocabulary if I have one: [VOCAB LIST OR NONE] </inputs> <task> Build a keyword list inside [VENUE LIMITS]. Prefer words that appear in [PASTE ABSTRACT] or [TERM LIST]. If I pasted [VOCAB LIST OR NONE], map my terms to that list only when the mapping is obvious from the words. Flag any keyword that would imply a literature I did not write. </task> <constraints> - Do not invent MeSH-style terms from memory if I did not paste that vocab. - Do not stuff keywords with journals or authors I did not cite. - Do not add a method I did not run. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Keyword list, source word in the paste for each, rejected words, UNVERIFIED rows. </format>

<context> I need a running head or short title that still names the same study. Do not invent a punchy label that changes the claim. </context> <inputs> - Abstract or full title: [PASTE ABSTRACT] - Current title: [WORKING TITLE] - Character limit: [CHARACTER LIMIT] - Venue rule: [VENUE LIMITS] </inputs> <task> Draft 5 running heads or short titles at or under [CHARACTER LIMIT]. Each must be a shortening of [WORKING TITLE] or of terms in [PASTE ABSTRACT], not a new topic. Count characters with spaces. Flag any shortening that drops a hedge or a population the title needs. </task> <constraints> - Do not invent a method or site name to look specific. - Do not replace a precise term with a trendier synonym that is not in the paste. - Honor [VENUE LIMITS] if they cap case or the words RUNNING HEAD. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Five short titles with character counts, drop-risk notes, UNVERIFIED rows. </format>

<context> I want the title to match the results I can defend. A title that names a finding I do not have is a honesty problem, not a branding problem. </context> <inputs> - Current title: [WORKING TITLE] - Results I can defend: [PASTE RESULTS] - Abstract if I have it: [PASTE ABSTRACT OR NONE] - Venue limits: [VENUE LIMITS] </inputs> <task> Audit [WORKING TITLE] against [PASTE RESULTS]. Tag the title MATCH, STRETCH, or NOT IN PASTE. If it names a method, population, or finding that is not in the results, rewrite 5 weaker titles I can keep. Do not add a paper that would save a loud title. </task> <constraints> - Do not invent results that would make the current title true. - Do not invent a colon subtitle with a theory I did not test. - Prefer a narrower true title over a punchy false one. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Verdict, five rewritten titles, words you dropped and why, UNVERIFIED rows. </format>

<context> I have a keyword list I am tempted to stuff. Audit it against the draft I wrote. Do not add a better keyword from a field I did not paste. </context> <inputs> - Keyword list: [KEYWORD LIST] - Draft or abstract: [PASTE DRAFT] - Venue limits: [VENUE LIMITS] - Terms I actually use: [TERM LIST] </inputs> <task> For each item in [KEYWORD LIST], tag KEEP (the draft supports it), STRETCH, or DROP. Propose replacements only from words already in [PASTE DRAFT] or [TERM LIST]. Do not fill the list up to [VENUE LIMITS] with extras I did not study. </task> <constraints> - Do not invent a controlled-vocabulary term from memory. - Do not add authors, journals, or datasets I did not cite. - Empty slots are better than trendy words. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Table: keyword, tag, replacement or DROP, UNVERIFIED. Then a clean list using KEEP rows only. </format>

<context> I must cut an abstract to [WORD LIMIT] without losing the argument or inventing tighter evidence. </context> <inputs> - Abstract: [PASTE ABSTRACT] - Target word count: [WORD LIMIT] - Must keep: [MUST KEEP] - Venue: [VENUE] - Citation rule: [NO CITES / ONLY PASTED CITES] </inputs> <task> Cut to [WORD LIMIT]. Protect [MUST KEEP]. Prefer deleting repetition, doubled examples, and throat-clearing. Do not delete a citation and keep the claim as if it were mine. Do not invent a tighter statistic to save space. Show a cut log. </task> <constraints> - Do not add new citations while cutting. - Do not merge two sources into one name. - Do not invent n or a percent that was not in the paste. - If the claim cannot survive the cut, say so and stop before you gut the evidence. - If a claim needs a source I did not give, mark it UNVERIFIED. - Report old and new word counts. </constraints> <format> Cut abstract, word counts, cut log (what left and why), UNVERIFIED rows. </format>

<context> My abstract verbs may overclaim. I want hedges that match notes I pasted, not decorative may on every line. </context> <inputs> - Abstract: [PASTE ABSTRACT] - Evidence notes or results: [PASTE EVIDENCE NOTES] - Work status: [COMPLETED / IN PROGRESS] </inputs> <task> Adjust claim strength in [PASTE ABSTRACT]: prove / show / suggest / may / cannot say. Tie each choice to a line in [PASTE EVIDENCE NOTES]. Do not add evidence. Do not hedge a count I recorded. If status is IN PROGRESS, refuse past-tense findings. </task> <constraints> - Do not invent a statistic that would let me drop a hedge. - Do not add citations. - Over-hedging ("it could perhaps be argued") is also a miss. Prefer one accurate verb. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Revised abstract with verbs marked, strength table (claim, verb, evidence line), UNVERIFIED rows. </format>

<context> I have reviewer, editor, or supervisor comments on an abstract. I will revise the abstract I paste. I will not invent sources to look responsive. </context> <inputs> - Abstract: [PASTE ABSTRACT] - Comments: [PASTE COMMENTS] - Notes I can use: [PASTE NOTES] - Word limit: [WORD LIMIT] </inputs> <task> Map each comment in [PASTE COMMENTS] to: agree and change, ask for clarification, or push back with evidence already in [PASTE NOTES]. Then draft the revised abstract within [WORD LIMIT]. If a comment demands a source I do not have, write FIND THIS SOURCE, do not invent it. </task> <constraints> - Do not add papers the comments "reminded you of." - Do not treat every comment as law if it contradicts the pasted evidence. Say so. - Do not claim a change you did not draft. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Comment table (comment, decision, draft change), revised abstract, FIND THIS SOURCE list, UNVERIFIED rows. </format>

<context> I need an honesty pass: does this abstract say only what my notes support? </context> <inputs> - Abstract: [PASTE ABSTRACT] - Notes, tables, or draft: [PASTE NOTES] - Work status: [COMPLETED / IN PROGRESS] </inputs> <task> For each claim in the abstract, tag MATCH (paste supports it), STRETCH (stronger than the paste), or NOT IN PASTE. Quote a short supporting phrase for MATCH. Do not fix STRETCH by inventing evidence. Propose a weaker wording I can use until I find a source. </task> <constraints> - Do not add sources. - Do not treat your own memory of a famous paper as evidence. - If the abstract cites a work that is not in [PASTE NOTES], flag CITE NOT IN PASTE. - If status is IN PROGRESS, tag past-tense results as STRETCH unless the paste shows they are done. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Claim table: tag, quoted support or none, weaker wording. Then CITE NOT IN PASTE and UNVERIFIED lists. </format>

<context> I need to know whether the abstract promises a cleaner paper than I wrote. Compare the two texts I paste. Do not invent the missing paragraph. </context> <inputs> - Abstract: [PASTE ABSTRACT] - Draft or near-full paper: [PASTE DRAFT] - Venue: [VENUE] - Word limit: [WORD LIMIT] </inputs> <task> Compare [PASTE ABSTRACT] to [PASTE DRAFT]. Flag: a result that is cleaner than the paper, a method the paper never states, a limit the abstract dropped, and a hedge the abstract lost. Then rewrite the abstract within [WORD LIMIT] so it matches the weaker, true paper. Do not fill holes with new findings. </task> <constraints> - Do not invent papers, data, or a paragraph the draft still lacks. - If the draft is missing methods or results, write NOT IN DRAFT and shrink the abstract. - Do not make the rewrite more positive than the discussion. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Mismatch list, rewritten abstract, word count, NOT IN DRAFT and UNVERIFIED rows. </format>