Writing Style

How this skill works

This SKILL.md ships as a neutral template inside the plugin marketplace. Your personalized voice override lives in your project at scribetronic/style/writing-style.md. Skills that need voice always look there first; this template is the fallback.

To create the override in your project:

scribetronic style    # copies this template to scribetronic/style/writing-style.md, then opens it

After that, every drafting/editing skill will read your override automatically.

If you don't have a scribetronic/ directory in the project (i.e. you installed the plugin without the CLI), the rules below apply directly — they're a generic baseline good enough to start, but you'll want to personalize.

Metadata

• status: template — fill in your specifics in scribetronic/style/writing-style.md
• last reviewed:

Writing style — <your-name>

The voice every post inherits. Format-specific skills (newsletter, hot-take, listicle, etc.) override structure and length, never voice. This file is yours to rewrite. Replace every section's placeholder text with your actual voice, your actual references, your actual examples.

0. Reference sources

The authors / posts / accounts whose voice you actively model. Not for citation — for the slop-checker and the drafting skills to know what "in-voice" sounds like.

• <reference 1> —
• <reference 2> —
• <reference 3> —

Own samples (your published work that defines the baseline — style-extract reads these):

• <your piece 1> —
• <your piece 2> —

1. Voice & tone

The 1-2 line answer to "what does this writer sound like?". Replace the defaults with your own.

Default starting point (delete and rewrite):

Direct, confident, honest. The reader should feel they're hearing from someone who's actually done the thing — not someone teaching from a book.

Tensions to maintain (keep the ones that fit; cut or replace the rest):

• Authoritative but not preachy. State the rule. Don't lecture about why it matters for three paragraphs.
• Personal but not confessional. Use "I" freely. Don't dump emotion.
• Educational but not flexing. Numbers are evidence, not trophies.
• Pragmatic over inspirational. "If it doesn't work, kill it" beats "embrace the journey".

Positive example (your own writing): <a 1-2 sentence quote from your own published work that is unmistakably you>

Negative example (your own writing): <a 1-2 sentence quote from a draft you later cut or regretted — the kind of writing you want to avoid>

2. Structure

How a post is shaped at the macro level. Adjust to your defaults.

• TL;DR or thesis up front. First 2 sentences must give the reader the point. If you can cut everything after the third paragraph and still keep the message, do it.
• Descriptive H2s, not clever ones. "What is X?" beats "Meet the new kid on the block."
• Short paragraphs. 1-3 sentences. Single-sentence paragraphs are punctuation.
• Bullets only when enumerating. Never to pad. Three bullets minimum or it's a list of 1-2 disguised as a list.
• Closers must close, not soften. Specific next action, declarative statement, or one-line punch. Never "let's see where it takes us."

3. Sentence-level

• Length variance is the engine. 4-12-20-6. Don't write three medium sentences in a row.
• Em-dashes over parentheses for asides. Cleaner rhythm. Reserve (parens) for genuine side-info that could be cut.
• Cut hedges. "I think", "probably", "kind of", "really", "truly", "honestly". One per post max — and only if it carries weight.
• Strong verbs, plain nouns. "Ship" not "deliver". "Tool" not "solution". "Built" not "created".

Native-language interference (delete if you write only in your native language):

• <your common leak 1> →
• <your common leak 2> →

4. Signature moves (keep these)

The deliberate techniques you want every post to use when they fit. Replace with yours.

1. <Move 1 — short name>. <1-2 sentence description>
2. <Move 2 — short name>.
3. <Move 3 — short name>.
4. Staccato 2-3 word paragraph. A single one-line sentence used as punctuation. "Why?" "That's it." Limit: ONE per post. More than that = tic.
5. Concrete artifact reference. Name the tool, the URL, the dollar amount, the date. Vague references signal you don't actually do the thing.

5. Anti-patterns / blacklist

Patterns to kill on sight. The defaults below catch the worst of generic AI output. Add your own.

6. Positive examples (from your own published work)

Pull 2-4 short quotes that are unmistakably your voice. Annotate why each one works.

Number-led opener: —

Honest aside: —

Punch closer: —

7. Negative examples (cut these on sight)

Pull 1-2 quotes from your own drafts (or published work you regret) that exemplify the anti-patterns above. Show the fix.

Before:

What's wrong: .

Fix:

8. Revision checklist

Run before publishing any post:

• First 2 sentences contain the point.
• Closer is declarative, not soft. (No "Let's see…")
• At most ONE rhetorical question.
• At most ONE staccato 2-3 word paragraph ("That's it.", "Why?").
• Zero "not just X, but Y" constructions.
• Zero adjective stacks of 3+.
• At least one concrete number, name, or date.
• No filler hedges (kind of, really, truly, honestly, probably) unless load-bearing.
• No coach-speak (journey, embark, transformational, unleash).
• Sentence length varies — read aloud test.
• If "I" appears in a feeling-claim, it's earned by a fact next to it.

If any check fails, revise. If 3+ fail, the draft isn't ready.

How to fill this file

Two paths:

1. Manual. Read each section, replace placeholders with your real voice, references, examples. Best when you already have a clear sense of your style. Budget: 30-60 minutes.
2. Assisted. Run /scribetronic:style-extract and feed it 3-5 reference samples (other authors you admire) plus 1-2 of your own published pieces. It generates a first draft of all 8 sections. Then you edit. Best when you're starting from scratch or want a structured baseline.

Refining this guide

This file should evolve as your published voice diverges from what's written here. Don't rewrite it from memory — run the /scribetronic:style-refine skill, which proposes concrete deltas based on your real (draft → published) edit history. You review, accept the deltas you agree with, and apply them by hand. No auto-rewrites.