write

Overview of write skill

write is an editing skill for rewriting prose in Chinese or English so it sounds natural, clear, and less AI-like without changing the core meaning. It is best for people who need a fast write skill install for drafts, docs, release notes, launch copy, social posts, or technical writing cleanup, especially when the real job is to preserve intent while removing generic phrasing and stiff structure.

What write is for

Use write when you already have text and want a better version, not when you need fresh ideation from zero. It fits users asking to rewrite, proofread, polish, cut AI tone, or tighten a passage for public reading. The main value is that it treats prose as a working artifact: readable, audience-aware, and specific.

Best-fit use cases

The strongest fit is editorial cleanup for Chinese and English content that already has substance but needs shaping. That includes product updates, blog drafts, launch notes, internal docs, and write for Technical Writing tasks where clarity matters more than style flourish. It is less useful for code comments, commit messages, or inline documentation.

Key differentiators

write is not a generic rewrite prompt. It ships with language-specific rules and repository references, so users can follow a repeatable write guide instead of guessing how to prompt each time. The skill also emphasizes removing formulaic AI patterns, not “improving vocabulary” for its own sake, which keeps edits closer to the original author voice.

How to Use write skill

Install and locate the workflow

Install with npx skills add tw93/Waza --skill write. After installation, start with SKILL.md, then read the supporting references in references/ before adapting the workflow to your own repo or writing environment. For most users, the fastest path is to preview the language-specific reference that matches the text you are editing.

What input the skill needs

write works best when you provide the actual text, the intended audience, and the format target. A weak request is: “make this better.” A stronger request is: “rewrite this Chinese product update for external readers, keep the facts, remove repetitive phrasing, and keep it concise.” If the audience is unclear, state it upfront so the rewrite does not drift into the wrong tone.

Good prompt shape

A practical write usage prompt should include four things: source text, audience, language, and constraints. For example: “Polish this English launch paragraph for a technical audience; keep the message, reduce filler, and make it sound human.” For bilingual content, say whether you want a translation review, a bilingual rewrite, or a pure prose edit, because the skill uses different reference paths for each.

What to read first

Start with SKILL.md to understand the operating rules, then inspect references/write-en.md, references/write-zh-prose.md, references/write-zh-bilingual.md, and references/write-zh-release-notes.md as needed. The reference files contain the real leverage: they show how the skill handles English prose, Chinese prose, bilingual consistency, and release-note style. That is more useful than treating write as a single universal prompt.

write skill FAQ

Is write only for polishing text?

No. The write skill is mainly for rewriting prose so it reads naturally and keeps the original intent intact. It can also be used for proofread-style cleanup, coherence fixes, and tone adjustment, but it should not invent missing content or expand a draft beyond what the source supports.

When should I not use write?

Do not use write if you have no source text, if the audience is unknown, or if the task is code-adjacent text like commit messages and inline comments. It is also a poor fit when you want heavy creative rewriting, marketing invention, or a complete structural redesign instead of editorial improvement.

How is this different from a normal prompt?

A normal prompt relies on your one-off instructions. write gives you a reusable write guide with preflight checks, language-specific references, and constraints that reduce guesswork. That usually leads to more consistent output, especially when you need the same editing standard across multiple documents.

Is it beginner-friendly?

Yes, if you can paste text and describe the target audience. Beginners get the best results when they keep the request narrow: “rewrite this paragraph naturally” works better than “make it great.” The skill is most helpful when users know what they want the text to do, even if they do not know how to explain the style technically.

How to Improve write skill

Give sharper source context

The biggest quality jump comes from better input. Tell write what the text is for, who will read it, and what must stay unchanged. If you are using write for Technical Writing, include the product or system context, the reader level, and any terms that should remain untouched so the rewrite does not flatten important detail.

Remove ambiguity before you ask

Common failure modes come from vague targets: unclear audience, mixed languages, or source text that is already overloaded with multiple purposes. If a paragraph is meant to inform and persuade, say which goal wins. If the text contains Chinese and English, say whether you want bilingual consistency or a single-language cleanup.

Iterate on the first draft

Treat the first output as a diagnostic pass. If it sounds too formal, ask for a lighter edit. If it loses precision, ask for a closer rewrite that preserves terminology and sentence meaning. The best write usage pattern is iterative: fix tone first, then tighten wording, then check whether any facts or audience cues were distorted.

Use the right reference path

If the output still feels off, improve the inputs by matching the reference file to the task instead of writing a broader prompt. Use references/write-en.md for English, references/write-zh-prose.md for Chinese prose, and the bilingual or release-note references when the format demands them. This is the main way to get better results from write install without overprompting.