prose
npx machina-cli add skill tslateman/duet/prose --openclawWriting Clearly and Concisely
Overview
William Strunk Jr.'s The Elements of Style (1918) teaches you to write clearly and cut ruthlessly.
WARNING: references/elements-of-style.md consumes ~12,000 tokens. Read it only when writing or editing prose.
When to Use This Skill
Use this skill whenever you write prose for humans:
- Documentation, README files, technical explanations
- Commit messages, pull request descriptions
- Error messages, UI copy, help text, comments
- Reports, summaries, or any explanation
- Editing to improve clarity
- Tightening or sharpening a draft that runs long
If you're writing sentences for a human to read, use this skill.
Limited Context Strategy
When context is tight:
- Write your draft using judgment
- Dispatch a subagent with your draft and
references/elements-of-style.md - Have the subagent copyedit and return the revision
All Rules
Elementary Rules of Usage (Grammar/Punctuation)
- Form possessive singular by adding 's
- Use comma after each term in series except last
- Enclose parenthetic expressions between commas
- Comma before conjunction introducing co-ordinate clause
- Don't join independent clauses by comma
- Don't break sentences in two
- Participial phrase at beginning refers to grammatical subject
Elementary Principles of Composition
- One paragraph per topic
- Begin paragraph with topic sentence
- Use active voice
- Put statements in positive form
- Use definite, specific, concrete language
- Omit needless words
- Avoid succession of loose sentences
- Express co-ordinate ideas in similar form
- Keep related words together
- Keep to one tense in summaries
- Place emphatic words at end of sentence
Section V: Words and Expressions Commonly Misused
Alphabetical reference for usage questions
Editing by Subtraction
Sharpening a draft means removing everything that does not serve the core point. Not rewriting. Not adding "voice." Cutting.
Core Principles
Cut Structure, Keep Substance. Drafts bloat through structure, not through ideas. Headers, transitions, framing paragraphs, and setup sentences add length without adding meaning. Cut them first. The substance survives.
Catch LLM-isms. AI-generated text carries recognizable signatures: contrastive pivots, filler authority phrases, uniform tone, and predictable word choices. These patterns weaken prose even when the underlying argument is strong. Identify them and cut or rewrite.
Substance Over Voice for High-Stakes Prose. "Humanizing" prompts add colloquialisms and personality but dilute the core argument. For leadership pitches, technical proposals, and executive communication, fluff hurts more than polish helps. Prioritize the argument.
Trust the Author's Cuts. A single note from the author ("cut 'The stakes are real'") often does more than an elaborate rewriting pass. The author knows what matters. When they cut, follow the instinct rather than restoring through rewording.
LLM-ism Catalog
Recognizable AI writing patterns and their fixes:
| Pattern | Example | Fix |
|---|---|---|
| Contrastive pivot | "That's not a trend piece -- that's the stakes" | State what it is. Skip the negation setup. |
| Needless stakes declaration | "The stakes are real." | Delete it. If the argument is strong, it shows. |
| Uniform sentence rhythm | Every paragraph same length, same cadence | Vary length. Let some sentences punch. |
| Hedged authority | "It's worth noting that..." | Drop the hedge. State the point. |
| Formulaic structure | Intro paragraph, three body sections, summary | Start with the strongest point. Cut the frame. |
| Predictable word choices | Always the most probable next word | Choose the specific word, not the expected one. |
| Throat-clearing openers | "When we think about X, we need to consider..." | Delete and start at the verb. |
Sharpening Workflow
- Read the draft. Identify the core point in one sentence
- Cut everything that does not serve it. Framing, transitions, restatements, setup paragraphs
- Scan for LLM patterns. Check the catalog above. Fix or delete each instance
- Verify nothing was lost. Read the shortened version. Does every important idea survive? If yes, the cut version is the better version
The Test
Can you remove this sentence without losing meaning? Remove it.
Bottom Line
Writing for humans? Read references/elements-of-style.md and apply the rules. Low on tokens? Dispatch a subagent to copyedit with the guide.
See Also
/naming— Strunk's Rule 12 applies identically to code names/review— Write review comments clearly using these rulesskills/FRAMEWORKS.md— Full framework indexRECIPE.md— Agent recipe for parallel sharpening (3 workers)
Overview
Apply Strunk Jr.'s The Elements of Style to prose across documentation, commits, error messages, and UI text. The skill helps you write with clarity and tighten drafts by subtraction, yielding a more professional voice. It emphasizes cutting superfluous parts to keep only the core point.
How This Skill Works
It applies elementary rules of usage and composition, then uses subtraction editing to trim bloat. Drafts are run through a copyedit pass (via a subagent) with references/elements-of-style.md, returning a tightened revision. Focus areas include active voice, definite language, elimination of needless words, and keeping related words together.
When to Use It
- Documentation, README files, and technical explanations
- Commit messages and pull request descriptions
- Error messages, UI copy, help text, and comments
- Reports, summaries, or any human-facing explanation
- Editing to improve clarity or tighten a long draft
Quick Start
- Step 1: Write a clear, human-readable draft.
- Step 2: Dispatch a subagent with your draft and references/elements-of-style.md for copyediting.
- Step 3: Review the revision and apply subtraction-based edits to tighten the prose.
Best Practices
- Use active voice
- Omit needless words
- Cut structure first; preserve substance
- Put statements in positive, definite language
- Keep related words together and maintain consistent tense
Example Use Cases
- Documentation: 'This document explains how to perform the operation.'
- Commit message: 'Add feature: improved functionality and performance.'
- Error message: 'Unable to complete. Please try again.'
- UI copy: 'Starting process…' instead of 'The system will attempt to start the process and keep trying until it finishes.'
- Report: 'We analyzed results to determine whether goals were met.'