agentleFS
georgejiehgeorgejieh/agent-skill-library-public/skills/writing/final-prose-review/SKILL.md

final-prose-review

Skill1 starsChanged 3 months agoMITView on GitHub

Use when prose needs a final pre-publication review.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not use as a substitute for domain review, legal review, fact checking, accessibility review, copy editing to a publisher's house style, …

    line 21
  2. Distinctive:

    Trace material claims: Verify names, dates, figures, quotations, causal language, links, and assertions that could alter a reader's …

    line 38
  3. Distinctive:

    Replace model-shaped prose: Remove topic announcements, empty significance claims, generic scene-setting, inflated verbs, vague …

    line 42
  4. Distinctive:

    Check coherence and navigation: Ensure headings describe what follows, transitions express real relationships, pronouns have clear …

    line 46
  5. Distinctive:

    Perform a restraint pass: Remove unsupported anecdotes, invented observations, false personal experience, fabricated quotations, ornamental …

    line 48
  6. Distinctive:

    Before handoff, confirm that:

    line 86
  7. Distinctive:

    every paragraph serves a distinct reader-facing purpose.

    line 90

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Final Prose Review

Use this skill after a draft is substantively complete and before human review, publication, or delivery. It is a final blind-spot pass for clarity, evidence, genre fit, and natural cadence. It improves the document itself; it does not claim to prove authorship, imitate a person, or evade a detector.

When to Use

  • A report, article, proposal, technical document, memo, research draft, or public page has a complete first draft.
  • A document reads as generic, over-produced, mechanically structured, vague, promotional, or insufficiently grounded.
  • A writer needs a final review that catches common model-generated prose failures before a human reviewer receives the document.

Do not use as a substitute for domain review, legal review, fact checking, accessibility review, copy editing to a publisher's house style, or consented voice calibration.

Prerequisites

  1. Identify the document's genre, audience, purpose, required action, publication context, and length or format limit.
  2. Gather the current draft, its source list, the governing style guide or template, and the author-approved facts, quotations, and terminology.
  3. Classify each material sentence as verified fact, attributed claim, analysis, recommendation, quotation, or unresolved statement.
  4. Preserve the original draft or a reviewable diff before revising.

Completion: the reviewer has the document contract and can distinguish source-backed content from prose that still needs verification.

Procedure

  1. Read for the reader's task. State in one sentence what the reader should understand, decide, or do after reading. Compare the title, opening, headings, and close against that task.
    • Completion: the document has one intelligible purpose, and its first screen tells the reader why it matters.
  2. Check genre before style. Apply the governing genre's factual, structural, attribution, citation, and format requirements. Preserve distinctions such as news versus analysis, explanation versus procedure, and research finding versus recommendation.
    • Completion: the form matches the reader's need rather than merely sounding polished.
  3. Trace material claims. Verify names, dates, figures, quotations, causal language, links, and assertions that could alter a reader's decision. Attribute claims to a source with the relevant authority and scope; label unresolved claims rather than smoothing them into certainty.
    • Completion: no material sentence is stronger, broader, or more current than its evidence.
  4. Audit the argument or action spine. For every paragraph, identify its job: answer, evidence, explanation, qualification, transition, or action. Delete, merge, or move paragraphs that repeat an earlier job without adding information.
    • Completion: each paragraph advances the reader's understanding or decision.
  5. Replace model-shaped prose. Remove topic announcements, empty significance claims, generic scene-setting, inflated verbs, vague attribution, rhetorical questions answered immediately, predictable three-part lists, and formulaic conclusion sections. Replace each with a specific fact, useful analysis, direct instruction, or deletion.
    • Completion: the draft contains information rather than performance of informativeness.
  6. Restore natural cadence. Vary sentence length where the meaning benefits, but do not add quirks for their own sake. Prefer concrete nouns, explicit actors, direct verbs, and precise repeated terms over decorative synonyms. Keep warranted qualifiers and remove stacked hedges.
    • Completion: the prose reads as deliberate and controlled when spoken aloud, without manufactured personality.
  7. Check coherence and navigation. Ensure headings describe what follows, transitions express real relationships, pronouns have clear referents, terms remain consistent, links identify their destination, and lists or tables aid scanning.
    • Completion: a reader can skim the document without losing the argument or action path.
  8. Perform a restraint pass. Remove unsupported anecdotes, invented observations, false personal experience, fabricated quotations, ornamental metaphors, unnecessary bolding, and calls to action that do not follow naturally from the document's purpose.
    • Completion: every surviving flourish earns its place through meaning, not decoration.
  9. Report remaining limits. Separate corrections made from issues that require a fact owner, subject-matter expert, lawyer, editor, or author decision.
    • Completion: the handoff distinguishes ready-for-human-review from unverified or out-of-scope material.

Decision Points

ConditionAction
A fact lacks a source or reliable ownerMark it unresolved, narrow it, or remove it.
The draft mixes news, analysis, and opinionLabel the transition or choose one primary form.
A sentence sounds polished but adds no informationReplace it with evidence or delete it.
A quotation is vivid but does not advance the pointParaphrase, shorten, or remove it.
A style rule conflicts with accessibility, accuracy, or a publisher requirementFollow the binding requirement and record the exception.
Revision would change the author's position, evidence, or intended claimFlag it for author approval instead of silently rewriting it.

Blind-Spot Checklist

  • Is the answer, request, or finding visible before background context?
  • Does every heading earn its place by helping a reader navigate?
  • Is each statistic paired with its source, timeframe, population, and material limitation?
  • Are claims of cause, novelty, scale, consensus, urgency, or impact supported at the stated strength?
  • Are sources named where attribution matters, rather than hidden behind vague collective phrases?
  • Has the draft avoided fabricated people, experiences, quotations, citations, and observations?
  • Does the ending provide a real conclusion or next action instead of restating that the topic matters?
  • Would a skeptical editor know which statements remain unverified?

Pitfalls

  • Detector chasing: Replacing phrases solely to manipulate a detector produces worse writing and does not establish authorship. Improve specificity, structure, and truthfulness instead.
  • Thesaurus drift: Replacing precise repeated terms with near-synonyms can obscure technical meaning.
  • False certainty: Removing every hedge can turn properly qualified evidence into an unsupported claim.
  • Uniform rhythm: Making every sentence equally short, equally long, or equally shaped produces mechanical prose.
  • Review overreach: A prose review may improve clarity but must not silently alter an author's viewpoint or an organization's position.
  • Cosmetic citation: A formatted reference does not prove that the source supports the sentence beside it.

Verification

Before handoff, confirm that:

  • the document's purpose, reader, genre, and required action remain explicit;
  • material claims, quotations, figures, names, dates, and links were checked or marked unresolved;
  • every paragraph serves a distinct reader-facing purpose;
  • removal of generic or inflated language did not erase warranted nuance;
  • the final document follows its binding template, style, and accessibility requirements; and
  • remaining factual, legal, editorial, or authorial decisions are listed plainly for human review.

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.