I Turned My 6 Best SEO Prompts Into Claude Code Skills. Now They Run Themselves.

Instead of copying the same prompts into new chats every week, I turned them into reusable skills for research, outlining, editing, internal links, snippets, and metadata.I pasted the same 400-word prompt weekly.My Notes app held six SEO prompts I copied into fresh chats, then fixed the formatting…

Instead of copying the same prompts into new chats every week, I turned them into reusable skills for research, outlining, editing, internal links, snippets, and metadata.I pasted the same 400-word prompt weekly.My Notes app held six SEO prompts I copied into fresh chats, then fixed the formatting each time. Now I type /serp-intent and Claude Code pulls that full prompt into my project folder. A skill is plain markdown with a name and a trigger line, stored under .claude/skills.Writers in one open write-up hit 80% first-pass readiness when structured context stayed loaded. Below you get my six files in full, plus the folder shape to run them yourself.Why Skills Beat a Cluttered Notes App: Quick OverviewSkills cut the copy-paste loop for good.Local run for speed: files sit in .claude/skills and load with a slash command, so you stay in the terminal instead of juggling chat tabs.Steady output at 80%: one write-up logged 80% first-pass readiness when structured context stayed in place through the draft.Plain text only: each skill needs a name plus one description line, with plain markdown pasted below and zero setup beyond the folder.The File Architecture: One Directory, Plain Markdown, Two Header LinesSix files sit in one hidden folder.My project root holds a folder named.claude/skills, and each skill gets its own subfolder with a SKILL.md file inside. The Claude Code prompt library documentation uses the same name plus description shape, so the pattern feels familiar fast.That file starts with two header lines, then your prompt text:---name: serp-intentdescription: Use when I need search intent from SERPs and rival posts before outlining---Paste your full prompt below the second --- line.Claude Code reads the description to decide when the skill fits. You call it with a slash, like /serp-intent best CRM for freelancers, and the text below the header loads as instructions.This visual run shows slash commands loading local files:You keep full control of voice because nothing runs in the background.Quick rule: Keep the description under 30 words and start it with Use when.If the terminal feels new, this guide to setting up claude code for non-programmers walks through folder basics for writers.Skill 1: /serp-intent for Searcher Need Extraction/serp-intent reads the top results first. I run this skill before I outline anything. It turns five rival URLs plus the search page into a tight intent map I can write against.How It WorksThis skill takes a keyword and a small set of URLs, then sorts what searchers want into buckets you can assign to headings.You are a search intent analyst. Inputs: KEYWORD = [paste keyword], URLS = [paste 4 to 6 rival URLs].1. Read each URL and list the job the reader hired it to do.2. Output three groups: Primary intent (the main job), Secondary intents (2 to 3 follow-up jobs), Fractured intent (where results split).3. For each intent, give: label, proof quote from the SERP, pages that serve it.4. Close with 5 reader questions the current top results answer poorly.5. Keep the full output under 350 words. Use plain lists.You paste the keyword and URLs after the slash command in the same line.Type /serp-intent plus your keyword in Claude Code.Paste 4 to 6 rival links when it asks for inputs.Copy the intent labels straight into your outline headings.What It ReplacesIt replaces open tabs and gut calls about angle. You get labels you can paste into headings instead of rereading the same five posts.ExampleA freelance writerlda tested this on the best budget microphone for podcasting. She pasted the keyword plus five review posts and got a split between setup help and pick lists.Weak: Write a general mic roundupStrong: Lead with setup steps for new podcasters, then rank three USB picksIt worked because the fractured intent showed setup questions outranked pure product chatter.Skill 2: /outline-brief for SERP-Driven Article ArchitectureOutlines drift without assigned jobs. I use this skill to turn the intent map into H2 and H3 order. Each heading leaves with one job and one question to answer.How It WorksThis skill builds heading order from your keyword, intent labels, and word aim, so drafting stays on rails.You are an outline builder for SEO drafts. Inputs: KEYWORD = [paste], INTENT MAP = [paste from /serp-intent], LENGTH = [paste word aim].1. Propose one H1 under 60 characters with the keyword near the front.2. Propose 5 to 7 H2s in search order. Under each H2, add 2 to 3 H3s.3. Under each H2, write: intent served, reader question answered, points to cover in 2 lines.4. Mark where a list, table, or short definition block fits.5. Flag any H2 that repeats intent from another H2 and merge it.You run it right after the intent skill while that map is still in context.Paste the intent output into the LENGTH prompt.Ask for H2 order tied to search order.Lock the H1 and H2s before you draft a word.Tip: Paste the locked outline at the top of your draft file so each session starts from the same plan.What It ReplacesIt replaces loose bullet outlines that wander by H3 number three. Every heading carries its assignment, so later sections stay tight.Skill 3: /voice-dna for Anti-Fluff Tone EnforcementFluff creeps back in long drafts.I run this skill on messy paragraphs to strip filler and flat rhythm. It holds my tone steady across a 2,000-word draft.How It WorksThis skill checks a pasted draft against a short voice profile, and returns edits you can accept fast.You are my line editor. VOICE PROFILE = [paste 6 to 10 lines: plain words, short sentences, direct address].1. Read the pasted draft section.2. List banned moves found: filler openers, passive verbs, stacked adjectives, repeated sentence length.3. Rewrite each flagged sentence in my voice profile. Keep meaning, cut 15% of words.4. Keep code, names, and numbers exact.5. Return: original line, edited line, one line saying why the edit holds voice.You feed it 300 to 500 words at a time for clean results.Paste your voice profile once per session.Paste one section of draft text.Accept line edits and rerun on the next section.For help building that profile, this guide to training AI on personal writing style shows a plain way to collect sample lines.What It ReplacesIt replaces slow self-edits where tired eyes miss filler. You get line pairs with reasons, so voice stays put from intro to close.Skill 4: /internal-links for Contextual Anchor MappingOld posts hide link chances. I point this skill at my draft plus a list of live URLs. It returns exact phrases where a link fits the sentence.How It WorksThis skill matches draft phrases to your URL list and suggests anchors that read as part of the sentence.You are an internal link mapper. Inputs: DRAFT = [paste markdown], URL LIST = [paste title + URL lines].1. Scan the draft for claims that a past post already covers.2. For each match, give: draft phrase (5 words or less), target URL, one sentence anchor that fits grammar.3. Allow 3 to 6 links per 1,500 words. Skip intro and conclusion.4. Never use click here, learn more, or read more as anchor.5. Return a list ordered by draft position with line numbers.You run it on markdown source so line numbers line up with your file.Export the draft as markdown.Paste a URL list of 20 to 40 live posts.Insert the top matches where the phrase already exists.What It ReplacesIt replaces manual site searches and forced links. You link from words you already wrote, so anchors sound natural.ExampleA blogger mapped a new SEO brief against 32 live URLs. The skill flagged setup steps and outline tips as link targets tied to exact lines.Weak: Read more about outlinesStrong: Link outline tips from the phrase lock the H1 and H2sIt worked because the anchor used words already in the draft instead of adding a generic prompt.Skill 5: /snippet-stealer for Position-Zero Answer BoxesGoogle lifts short answers first.I run this skill on finished H2s to add tight definition blocks and lists. Those blocks feed answer boxes, and People Also Ask picks.How It WorksThis skill rewrites each H2 core answer into a short block that can stand alone in results.You are a snippet editor. Inputs: H2 TEXT = [paste one H2 plus 200 words].1. Write one definition answer of 35 to 45 words. Put the term in the first sentence.2. If steps fit, add a 4 to 6 item numbered list with verbs first.3. If contrast fits, add a 3 row table with plain headers.4. Keep sentences under 20 words. Keep numbers exact.5. Return the block plus the H2 it belongs under.You process one H2 at a time to keep answers crisp.Paste a single H2 section.Pick the definition, list, or table form it suggests.Paste the block right under that H2 in your draft.What It ReplacesIt replaces long intros under each heading that bury the answer. Readers and result pages get the point in the first lines.Skill 6: /meta-polish for High-CTR Headline GenerationTitles decide the click.I run this skill last, once the draft is set. It gives me ten title angles plus meta lines that fit length caps.How It WorksThis skill reads your H1, outline, and main takeaway, then writes options you can test.You are a headline writer for SEO. Inputs: H1 = [paste], OUTLINE = [paste H2s], TAKEAWAY = [paste one line].1. Write 10 title options under 60 characters. Include the keyword in 7 of them.2. Cover 10 angles: how-to, number, cost, time, mistake, checklist, comparison, beginner, advanced, question.3. Write 3 meta descriptions of 140 to 155 characters with the keyword and one clear benefit.4. Rank your top 3 titles with one line each on why they pull clicks.5. No hype words. No extra punctuation.You pick one title and one description, then ship.Paste H1 plus H2 list.Ask for ten angles in one pass.Test the top three in your CMS preview.What It ReplacesIt replaces late-night title tweaks and truncated descriptions. You get set lengths and varied angles in one pass.Running the Chain: How to Assemble the Entire Publishing WorkflowOrder matters more than speed.I run the six skills in one line from keyword to meta. Each step feeds the next, so context stays narrow and clean.My chain looks like this in practice.Run /serp-intent on the keyword plus rival URLs.Run /outline-brief with that intent map.Draft to the locked H2s in your editor.Run /voice-dna on 500-word chunks.Run /internal-links on the markdown draft.Run /snippet-stealer, then /meta-polish last.The full set lives as plain files you can copy. Grab the starter shapes in this public GitHub Gist prompt skill library and compare patterns with this social media content skills repository, which holds 13 modular skills in the open.Small teams pick order by bottleneck.If drafts wander, then run intent plus outline back-to-back before writing.If voice slips, then run tone checks per section instead of at the end.If clicks lag, then run snippet plus meta passes before you hit publish.Use this pass checklist before you ship.Intent labels sit in H2 orderEach H2 answers one reader questionAnchors use exact draft phrasesSnippet blocks sit under target H2sTitle and description meet length capsFrequently Asked Questions About Writing With Claude Code SkillsDo I Need to Know How to Code to Run These Skills? You need terminal basics only. You make folders, paste markdown, and type slash commands.Can These Skills Access My Draft Files Directly? Yes, when you run Claude Code in your project folder. It reads markdown source and writes back suggested blocks.How Do I Edit or Update a Skill When My Editorial Preferences Change? Open the SKILL.md file and edit the text. Save, then call the slash command again to use the new version.Will These Skills Work on Standard Claude or ChatGPT Web Chats? You can paste the prompt text into any chat window. You lose the slash trigger and local file reads, but the words still work.This story is published under the Generative AI publication. Connect with us on LinkedIn and follow Zeniteq to stay in the loop with the latest AI stories. Let’s shape the future of AI together!I Turned My 6 Best SEO Prompts Into Claude Code Skills. Now They Run Themselves. was originally published in Generative AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

Source: Generative AI Pub — Published September 16, 2026 at 12:03 UTC — Category: Image AI