Four Skills earn their slot in a content marketer's setup: brief-to-draft, repurposing, editorial standards, and brand voice. Here is what goes in each, and the refuse rules that stop the slop.
Content marketing is the discipline most likely to drown in its own output. A blog post, three social cuts, a newsletter, a sales-enablement one-pager, all from the same source idea, all needing to sound like the brand, all due this week. That is exactly the shape of work Claude Skills are good at, because the standards are stable even when the topics change. You write the standard once and it loads on every piece.
This post is the content-marketing branch of the complete guide to Claude Skills for marketing. It covers the four Skills worth building, what goes in each (the trigger, what it knows, what it refuses), a full worked SKILL.md, and the refuse rules that matter most for content: refusing to pad word count and refusing to invent stats.
The temptation is a single "Content Skill" holding briefs, drafts, repurposing, and voice. It loads on everything, bloats the context, and produces worse output than a plain prompt. Build four narrow Skills instead, each with a tight trigger.
Each Skill below is defined by three things: the trigger (the description that tells Claude when to load it), what it knows (the reference files it points at), and what it refuses (the lines it will not cross). Same three-part shape used in the 20-minute build walkthrough.
Trigger: any request to turn a content brief into a first draft, or to outline a piece from a target keyword. Triggers on "draft this," "write the post," "turn this brief into," "outline a piece on."
Knows: the structure your team drafts in (H2 cadence, intro length, how internal links get placed) and the brief format you actually use. Pair it with a real brief from the free SERP-Informed Brief so the Skill reads the search intent the brief already captured rather than guessing.
Refuses: drafting a section the brief did not ask for, inventing a statistic to fill a claim, or padding the intro to hit a word count. If the brief is thin, it flags the gap instead of writing around it.
Trigger: any request to cut a long piece into other formats. Triggers on "repurpose this," "pull social out of," "turn this into a newsletter," "give me the LinkedIn version."
Knows: the formats you ship to and the constraints of each (the character ceiling on a given platform, the newsletter's section order, the one-pager template). It knows that a repurpose is a re-cut of existing material, not a fresh draft.
Refuses: adding new claims that were not in the source piece. This is the refuse rule that keeps repurposing honest. A social cut should not contain a number the article never made. If the source did not say it, the cut does not say it.
Trigger: any drafting, editing, or reviewing of content. This one runs wide on purpose, because standards apply to everything. Triggers on "draft," "edit," "review this," "proof," "tighten."
Knows: your style decisions. Sentence-case headings or title-case. Oxford comma or not. The words you ban (the "leverage" and "best-in-class" list). How you handle numbers, dates, and product names. This is your house style guide turned into a file Claude reads.
Refuses: the two refusals that matter most in content, covered in full below. No padding to length. No invented stats.
Trigger: any external-facing writing where it needs to sound like the brand. Triggers on "in our voice," "on-brand," "write," "rewrite."
Knows: the brand voice guide. The rhythm, the do/don't pairs, the register defaults. This is the one Skill almost every marketer should build first, and it is the one the build walkthrough uses as its worked example.
Refuses: inventing voice attributes not in the guide, and "punching up" the voice into something the brand is not. The voice is the voice.
Here is the editorial-standards Skill in full. It is short on purpose. A Skill that runs on every draft has to stay tight, because it is loaded constantly. The refuse rules are doing the heavy lifting.
--- name: editorial-standards description: Use this Skill whenever the user is drafting, editing, reviewing, or proofing any content piece (blog posts, newsletters, social, landing copy, one-pagers). Triggers on "draft," "edit," "review this," "proof," "tighten," "write the post," "turn this brief into." Loads the house style guide and enforces the two hard refusals: no padding to length, no invented statistics. --- # Editorial Standards ## What this Skill knows House style lives in style-guide.md (this folder). The decisions it enforces: - Headings in sentence case. One idea per heading. - Short sentences. Cut hedging. No throat-clearing intros. - Banned words list is in style-guide.md. Honor it. - Numbers: only use a figure if the source supports it. ## What this Skill does When loaded, apply style to every piece of output: - Match the structure in style-guide.md - Flag any sentence that exists only to add length - Keep the draft to the length the content earns, not a target ## What this Skill refuses - Refuse to pad word count. If a 900-word brief only has 600 words of real content, write 600 and say so. Do not invent sections, restate points, or stretch sentences to reach a number. - Refuse to invent statistics, percentages, dates, or named sources. If a claim needs a stat and none was provided, write "[stat needed: ...]" and move on. Never fabricate a number to make a sentence land. - Refuse to add a citation, study, or quote that was not in the source material. - If the user asks for filler to hit a length, say no and explain what is actually missing instead.
Most content Skills get written as "do" lists: do match the voice, do use the structure. The refuse section is where the value sits, because the two failure modes that hurt content most are both things Claude will do by default unless you stop it.
Refuse to pad word count. Ask for a 1,200-word post and a model will give you 1,200 words whether the idea is worth 1,200 words or not. The padding is invisible until a reader hits it: the restated intro, the section that just rephrases the last one, the sentence that says nothing. A Skill that refuses to pad produces a shorter, denser draft and tells you when the brief was thin. That is more useful than a draft that hides the gap. Length should be earned, not targeted.
Refuse to invent stats. This is the one that gets people in trouble. A model asked to make a claim sound credible will reach for a precise-looking number. "73% of marketers say." It reads well and it is made up. A content Skill that refuses to fabricate figures, and writes a [stat needed] placeholder instead, turns a publishing risk into a research to-do. The same discipline runs through the guide to SEO content briefs with Claude, where the brief flags which claims need a real source before drafting starts.
A Skill is only as good as the file it points at. The editorial-standards Skill enforces style-guide.md, but style-guide.md is the actual work. A vague style guide makes a vague Skill. Write the real one first, then wrap it.
Same for the brief-to-draft Skill: it is only as good as the brief. A SERP-informed brief in, a usable draft out. A thin brief in, and the Skill correctly refuses to fill the gaps for you.
If you build one this week, build the brand-voice Skill. It pays back on the first piece and it pairs with a free prompt. Then add editorial standards, the one that runs on everything and catches the padding and the invented stats. Repurposing and brief-to-draft come next, once the first two are stable. Three to five narrow Skills, each earning its trigger, beats one bloated Skill every time.
The four free Skills on the Hub are the fastest way to see the format. Open one, read the SKILL.md, and you have a working template.
The four free Claude Skills on the Hub are the same four free prompts packaged as installable Skills, with the YAML and the refuse rules already wired. Voice Extractor, Competitive Teardown, SERP-Informed Brief, Performance Readout. CC BY 4.0. No email gate. The best way to learn the format is to read one.
Get the 4 free Skills