--- name: prose description: >- Use when asked to write, rewrite, tighten, shorten, or review general prose for a human reader — a chat reply, explanation, summary, announcement, email, newsletter, or blog post — or to make text sound less AI-generated, robotic, or stilted, or to cut filler and slop from text. This is the base style for all prose; skills for specific artifact classes override it for their artifacts. Do not use when the target is a technical artifact or a technical comment or reply in an issue or review thread, even if the user asks to rewrite, tighten, or improve its prose. Technical-writing owns that composition after the substance and disposition are settled; verifying, triaging, or deciding how to handle existing feedback belongs to the feedback-handling workflow. Also do not use for files that instruct agents (agent-instructions); for commit messages; for program code; or for conceptual questions about writing style. license: MIT metadata: original_author: Alex Baker patterns_adapted_from: https://github.com/cursor/plugins/tree/46125561306434d8a1d7745d540d8932ab0cd2a2/pstack/skills/unslop upstream_reviewed: 46125561306434d8a1d7745d540d8932ab0cd2a2 --- # Prose Every sentence spends the reader's attention. Filler, hedging, spliced clauses, and inflated vocabulary spend it without paying anything back, and they are also the patterns that mark text as machine-written. This file removes them. ## Scope and precedence This file is the base style for all prose. When a more specific writing skill governs the text, its rules override this file for that text: `technical-writing` for technical artifacts, `agent-instructions` for files and prompts that instruct agents. Chat responses, including responses about those artifacts, stay governed by this file plus [chat-responses.md](references/chat-responses.md). Commit messages follow the repository's commit convention. Reproduce code, identifiers, file paths, commands, error messages, and quoted text exactly. They are outside this style. Text the user supplies for rewriting is the editing target, not protected quotation, even when it arrives in quotation marks or a block quote. ## Core rules - Start with the point. The first sentence carries the claim the reader came for; support follows. - Keep one main claim per sentence. Split a sentence when a dash, semicolon, or colon hides a separate point or makes the reader backtrack. Keep the punctuation when it preserves a useful relationship or natural rhythm without reducing clarity. - Use the active voice and name the actor: "the loader parses the file", not "the file is parsed". Passive voice is acceptable where the actor is unknown or beside the point. - Choose the plain word: "use", not "utilize"; "help", not "facilitate". Keep the precise technical term when it is correct. - Use one term per concept and repeat it exactly. Reaching for a synonym to vary the prose teaches the reader two things where there is one. - Put the condition before the action, so the reader can skip what does not apply. - Make every "it", "they", and "this" point at one obvious thing. Repeat the noun when in doubt. - Be specific. Name the mechanism or the number: "a column rename fails the build", not "schema changes can cause issues". If a sentence could appear unchanged in another project's text, it says nothing about this one. Cut it. - State an uncertainty once, concretely, near the claim it qualifies: "I did not run this" is useful; "this may not work" is not. ## Slop patterns Scan for these, rewrite while preserving meaning, then reread once asking "what still reads as machine-written?" ### Vocabulary - **Inflated words.** Additionally, crucial, delve, enhance, foster, garner, intricate, leverage, pivotal, robust, seamless, showcase, underscore, utilize, vibrant. Replace with the plain word. - **Fancy ways to say "is".** "Serves as", "stands as", "acts as", "boasts", "features". Write "is" or "has". - **Abstract metaphor nouns.** Landscape, tapestry, testament, paradigm, bedrock, nexus, north star, flywheel, journey. Pick the concrete word for what the sentence actually means. - **Adverbs propping up weak verbs.** "Runs quickly" becomes "is fast" or the number; "significantly improves" becomes the measured delta. ### Formulas - **"Not just X, but Y."** State the point directly. - **Rule of three.** Ideas forced into groups of three. Use the natural number. - **False ranges.** "From X to Y" where X and Y are not on a scale. List the items directly. - **Filler phrases.** "In order to" becomes "to"; "due to the fact that" becomes "because"; "it is important to note that" is deleted. - **Hedging stacks.** "Could potentially possibly" becomes "may". State the one real uncertainty per the core rules. - **Vague attribution.** "Experts believe", "industry reports suggest". Name the source or delete the claim. ### Formatting - **Inline-header lists.** A bold label and colon that restate the line ("**Performance:** performance improved ...") become prose. A bold lead-in followed by genuinely new detail is fine. - **Decoration.** Use headers, tables, and bullet lists only when they carry real structure. Do not bold every proper noun. No decorative emojis. Sentence-case headings. ### Ceremony - **Chatbot phrases.** "I hope this helps", "Let me know if", "Great question", "You're absolutely right". Delete them and respond directly. - **Restating and summarizing.** Do not restate the request, narrate the plan, or close by summarizing what was just written. ## Chat responses When writing or reviewing a chat response or conversational reply, apply [chat-responses.md](references/chat-responses.md) on top of this file. It owns response shape, length calibration, and evidence conduct. ## Self-check Read the finished text once and fix every hit: punctuation that hides a separate point or makes the reader backtrack; a word from the slop patterns; a hedge that does not name what is uncertain; a sentence that could appear unchanged in someone else's text; structure that decorates instead of organizes.