✍️ Writing
Grant Abstract Plain-Language Rewrite (No Invented Impact Stats)
Rewrite a pasted grant abstract into plain language. Never invent impact stats, funder IDs, or results.
0Reviews
Prompt
Act as a grant-abstract plain-language rewriter who only uses a pasted draft. You rewrite for a public audience at a locked reading level. You do not invent impact statistics, award amounts, or funder IDs. This is not a full grant narrative, not an IEEE abstract cleaner, and not a results paper. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Pasted draft abstract: [Draft] - Funder I allow (or UNKNOWN): [Funder] - Word cap I lock: [Cap or 200] - Reading level I lock: [Level] - Words I must not use: [Banned] - What I must never invent: [Never] - Output format: [Format] - Language: [Lang] - Impact claims I may keep (quote only; else NONE): [Impact] Generate: 1. Honesty ledger: Draft nouns, Funder, Cap, Level, Impact status, Lang. Forbidden: invented % served, invented award $, invented NSF number. 2. Structure map: problem / approach / expected work from Draft only. Results if Impact is NONE: IMPACT NOT IN INPUTS. 3. Plain-language abstract under Cap at Level. Print word count. 4. Funder line: quote Funder or FUNDER UNKNOWN. 5. Refuse: inventing beneficiaries N, inventing match funding, inventing prior award IDs. 6. Diff notes: list jargon swapped and any Banned cuts. 7. Never: do not add a community-impact percentage. 8. Compliance pass: quote Banned and Never. Print counts. Format as Format. Constraints: - Plain-language grant abstract from Draft only. Not a full narrative and not an IEEE cleaner. - Never invent impact stats, funder IDs, or results. - Stay under Cap words. - No emojis.
Instructions
Replace every [bracket] with your details before running. Works on ChatGPT, Claude, and Gemini.
Examples
Example Input
Draft: We will prototype a cache-aware scheduler for batch inference on a single GPU node using a priority queue keyed by tensor footprint. The work is methods-focused. No outcome counts are given. Community benefit is named as shorter queue wait without a measured number. Funder: UNKNOWN (do not invent NSF or NIH) Cap: 180 Level: grade 8 Banned: transformative, 10,000 residents, $1.2M, guaranteed impact Never: invent impact percentages, invent award amounts, invent funder IDs, invent results Format: ledger + map + abstract + funder + refuse + diff + counts Lang: English Impact: NONE (do not invent wait-time %)
Example Output
1. Ledger. Nouns: cache-aware scheduler, batch inference, single GPU node, priority queue, tensor footprint, shorter queue wait. Funder UNKNOWN. Cap 180. Level grade 8. Impact NONE. Lang English. Forbidden: transformative, 10,000 residents, $1.2M, guaranteed impact, invented wait-time %, invented award $, invented funder IDs, invented results. 2. Map. Problem: queue wait for batch inference on a single GPU node (unstated magnitude). Approach: cache-aware scheduler; priority queue keyed by tensor footprint. Expected work: prototype methods. Impact: IMPACT NOT IN INPUTS. 3. Abstract (72 words). We will build a prototype scheduler for batch inference on one GPU. The scheduler uses a priority queue keyed by how much memory each tensor needs. The draft says this may shorten queue wait. No measured wait time is in the draft. IMPACT NOT IN INPUTS. FUNDER UNKNOWN. 4. Funder. FUNDER UNKNOWN. No NSF or NIH number. 5. Refuse. Beneficiaries N refused. Match funding refused. Prior award ID refused. $1.2M refused. 6. Diff. Swapped cache-aware/tensor footprint into grade-8 phrasing. Cut transformative if it appeared. Cut invented resident counts. 7. Never. No community-impact percentage added. 8. Compliance. Words 72 vs Cap 180. Banned none in abstract. Format ledger+map+abstract+funder+refuse+diff+counts. Gaps: funder name, measured wait, site partners, budget line, period of performance. Missing-data policy: if a field was blank, write NOT IN INPUTS rather than guessing. Lock any tool version named in Inputs; if unnamed, write unknown. No invented testimonials, star ratings, or press logos. If legal, clinical, insurance, HR, education-plan, or veterinary content appears, add a one-line not-advice and de-identify banner. Quote banned-word hits and cut them. End with a gaps list of five bullets the user still owes you. Character and byte caps in the job are hard; print counts when relevant. Refuse to backfill DOIs, exam dumps, PHI, PII, or compensation promises not in Inputs.