agentleFS
megandmartinmegandmartin/agent-skills-repo/skills/content-marketing/seo-blog-writer/SKILL.md

seo-blog-writer

Skill4 starsChanged 4 months agoMITView on GitHub

Takes a keyword or content brief through search-intent analysis, an H2/H3 outline, and a full draft with E-E-A-T signals (first-hand experience, named sources), plus a meta title/description pair — no keyword stuffing, ever. Use when the user asks to "write a blog post", "SEO article", "rank for [keyword]", "draft from this content brief", or "outline a post". Don't use for landing/sales page copy — that's landing-page-copywriter.

8 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

SEO Blog Writer

Produces posts that earn rankings the durable way: match the searcher's actual intent before writing a word, structure with scannable H2/H3s, prove experience and cite real sources, and write for a human who might link to it. The keyword appears where it belongs and nowhere else.

When to Use

  • The user has a target keyword, topic, or content brief and wants an outline or full draft.
  • An existing post needs an intent-match or E-E-A-T upgrade.
  • The user asks for "meta title and description", "blog outline", or "article that ranks".
  • Not for: conversion-focused landing or sales pages (landing-page-copywriter), email content (email-nurture-sequence), or the underlying research when claims need sourcing first (deep-research-brief).

Quick Reference

ActionRule
Intent checkClassify the keyword: informational / commercial / transactional / navigational — before outlining
StructureOne H1; H2s answer the sub-questions searchers have; H3s only under a parent H2
Keyword useTitle, first 100 words, one H2, conclusion — then only where natural
E-E-A-T≥1 first-hand experience marker; ≥2 named, linkable sources; author-relevant framing
Density guardpython3 count in step 6 — exact keyword >1.5% of words = rewrite
Meta pairTitle ≤60 chars with keyword front-loaded; description 140–155 chars with a reason to click

Procedure

  1. Classify intent — from the keyword (and SERP notes if the user can share them), state the intent class and the searcher's real question in one sentence: "Someone typing X wants Y." If the brief's angle contradicts the intent (e.g., a sales pitch against an informational query), flag it and align before outlining — a mismatched post won't rank regardless of quality.
  2. Gather E-E-A-T raw material — ask the user for: first-hand experience with the topic (what they've done, built, tested, measured), 2+ credible sources they trust, and who the author is. No experience available → write honestly from research and say so; never fabricate "when I tested this".
  3. Outline — H1 (promise matching intent), then 4–8 H2s each answering one sub-question, H3s where a section splits. Include where the experience proof and each source lands. Deliver the outline for approval when the user asked for outline-first; otherwise proceed.
  4. Draft — answer the searcher's question in the first 100 words (no "in today's digital landscape" wind-ups), then earn depth section by section. Short paragraphs (≤4 lines), specifics over abstractions, every claim either sourced, experienced, or clearly framed as opinion.
  5. Weave E-E-A-T — experience markers in concrete form ("we ran this for 6 weeks; here's the number"), sources cited inline with name + link placeholder, and an author-context line if the platform shows bylines.
  6. Stuffing audit — count it, don't feel it: python3 -c "t=open('/tmp/draft.md').read().lower();w=t.split();k='KEYWORD';print(t.count(k), len(w), round(t.count(k)/max(len(w),1)*100,2))". Exact-match density above ~1.5%, or any sentence that exists only to hold the keyword, gets rewritten with variants or pronouns.
  7. Meta pair + deliver — write meta title (≤60 chars, keyword early, promise clear) and meta description (140–155 chars, states the payoff, gives a click reason). Assemble the Output Template.

Output Template

# [H1]

**Target keyword:** [kw] | **Intent:** [class] — "[searcher's real question]"
**Meta title (≤60):** ...
**Meta description (140–155):** ...

[Draft: answer-first intro → H2/H3 body with inline sources → conclusion + next step]

---
**E-E-A-T ledger**
- Experience markers: [where + what]
- Sources: 1) [Name — link] 2) ...
- Author context: [line or "n/a"]
**Density check:** [count]/[words] = [x.x]%

Pitfalls

  • Intent mismatch — a beautiful how-to guide targeting a keyword where searchers want a comparison. Recovery: redo step 1 honestly; restructure the outline to the dominant intent, keep the off-intent material as a section, not the spine.
  • Fabricated experience — "when we tested this" appears with no test behind it. Recovery: strip every experience claim the user didn't actually confirm in step 2; replace with sourced third-party evidence or cut.
  • Keyword stuffing by enthusiasm — the exact phrase lands in every H2 and image alt. Recovery: run the step 6 count; rewrite flagged sentences using natural variants; keep exact match only in the Quick Reference's four positions.
  • Wall-of-text sections — 300-word paragraphs under a single H2. Recovery: split any H2 section >250 words with H3s or a list; re-scan for a ≤4-line paragraph rhythm.
  • Generic AI intro — "In today's fast-paced world..." Recovery: delete the first paragraph; if the second paragraph answers the question, it's your new intro.

Verification

  • Intent class stated and outline visibly serves it
  • First 100 words answer the searcher's question
  • ≥1 real experience marker (user-confirmed) and ≥2 named sources with link placeholders
  • Step 6 density check run and printed in the ledger; ≤1.5% exact match
  • Meta title ≤60 chars, description 140–155 chars (character-counted, not guessed)
  • No sentence exists solely to contain the keyword
  • E-E-A-T ledger printed at the bottom of the deliverable (experience markers, sources, author context)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.