sciwrite prose

Scientific writing often suffers from dense, circular prose that obscures research value. This skill applies the principles of Stanford's Writing in the Sciences to tighten technical manuscripts, grant proposals, and abstracts at the sentence and paragraph level. It prioritizes clarity and the reader's effort over formal fluff.

What it does

• Prose auditing by identifying clutter, dead-weight openers, and redundant prepositions.
• Verb optimization by converting unnecessary passive voice to active and identifying buried predicates.
• Paragraph restructuring using the upshot method to eliminate repetition and logical gaps.
• Consistency checking for terminology, acronym usage, and numeric data across sections.
• Punctuation refinement using colons, semicolons, and dashes to vary rhythm and improve flow.

How it works

1. Submit text such as a single sentence, a full paragraph, or a file path for a manuscript section.
2. Execute scripts like prose_audit.py or keyword_consistency.py to flag candidates for revision.
3. Review the upshot for paragraphs, where the agent reduces text to plain-line points before rewriting.
4. Apply the five-pass audit (Clutter, Verbs, Punctuation, Terminology, Consistency) in sequence.

Frameworks & tools

This skill includes Python scripts for automated text extraction and auditing. It supports .tex, .docx, .md, and .txt formats.

Why this beats prompting it yourself

Generic prompts often result in "flowery" AI writing that violates scientific standards. This skill uses a specific five-pass protocol and the upshot method to ensure edits stay grounded in the Stanford methodology without changing scientific claims.

Use cases

• Shorten abstracts to meet strict word limits without losing critical data.
• Clarify Methods sections by fixing parallelism and terminology drift.
• Tighten Discussion sections by removing buried predicates and weak "to be" verbs.
• Polish grant proposals to ensure the main point and significance appear early.

Known limitations

Does not judge scientific correctness or restructure whole manuscripts into IMRaD sections. The Python scripts require local installation to run automatically on files.