Skill stacks for blog articles that rank well on Google

• npx skills add [owner/repo] without the -g flag installs at project scope (the current folder), which is the behavior we want.
• On the git clone routes, the destination is the .claude/skills/ folder inside the project (and .claude/agents/ for subagents).
• The stacks coexist with no conflict in the same .claude/skills/ folder, so combining them works.

Purpose: continuous production of optimized articles, with quality control by score. It is the most complete stack in the catalog, covering strategic planning through post-publication.

The stack's skills

Installation

npx skills add AgriciDaniel/claude-blog
npx skills add hardikpandya/stop-slop
git clone https://github.com/mackswendhell/humanizer-pt-br.git .claude/skills/humanizer-pt-br

Step by step — one-time project setup

1. /blog brand init: generates BRAND.md and VOICE.md (brand positioning and voice); every sub-skill loads those files on its own.
2. /blog persona: defines tone (NNGroup framework), readability and vocabulary of the editorial voice.
3. /blog strategy [your niche]: a plan with 20+ topics in hub-and-spoke clusters and a suggested calendar.

Step by step — per article

1. /blog brief [topic]: primary keyword + LSI, personas, outline, 5 recommended statistics.
2. /blog outline [topic]: H2/H3 structure based on the top 10 SERP, with the content gap identified.
3. /blog write [topic]: the full article: meta title and description, Key Takeaways, answer-first, sourced statistics, a hero image, SVG charts, FAQ and JSON-LD schema.
4. /blog factcheck [file]: checks whether each statistic appears in the cited source at all (an invented statistic destroys E-E-A-T).
5. /blog analyze [file]: a 0–100 score across 5 categories; iterate to 90+.
6. Humanization: Humanize este texto: [article] (Humanizer PT-BR) and then the Stop Slop score across the 5 dimensions (≥35/50). ⚠️ Do it before the technical validation: the rewrite can alter headings.
7. Technical validation: /blog seo-check [file] → /blog schema [file] → /blog geo [file] (on-page checklist, validated JSON-LD, AI citability score).
8. Post-publication: /blog google (Search Console, PageSpeed, indexing) and /blog repurpose [file] (X thread, LinkedIn, newsletter).
9. Monthly routine: /blog audit + /blog cannibalization (decaying content and posts competing for the same keyword).

Purpose: evergreen articles with original data, the kind of content Google rewards most (information gain) and that draws backlinks on its own. The research comes before the writing, with human approval of the scope before spending tokens.

The stack's skills

Installation

# Deep Research Skills (skills + web search subagent)
git clone https://github.com/Weizhena/deep-research-skills.git /tmp/drs
cp -r /tmp/drs/skills/research-en/* .claude/skills/
cp /tmp/drs/agents/web-search-agent.md .claude/agents/
cp -r /tmp/drs/agents/web-search-modules .claude/agents/
pip install pyyaml

# Writing and humanization
npx skills add AgriciDaniel/claude-blog
git clone https://github.com/mackswendhell/humanizer-pt-br.git .claude/skills/humanizer-pt-br

Step by step

1. /research [article topic]: generates the research outline: items to investigate + data fields (price, date, source…).
2. Review and edit the outline by hand: this is the human-in-the-loop moment. Adjust with: - /research-add-items: new angles of investigation; - /research-add-fields: new data columns (for example, "primary source", useful for the factcheck later).
3. /research-deep: dispatches one web search subagent per item, in parallel, filling in the data.
4. /research-report: compiles the markdown report. That report is your proprietary asset: data the competition lacks.
5. /blog brief [topic] and /blog write [topic], instructing Claude to use the report as a Tier 1 source, so the article comes out with exclusive data cited rather than generic statistics.
6. /blog factcheck → Humanizer PT-BR → /blog seo-check → /blog geo.

Purpose: when the blog exists to sell a product or service, the article has to rank and convert. Each text rests on the product's positioning, with a CRO and AI-SEO layer.

The stack's skills

Installation

npx skills add coreyhaines31/marketingskills
npx skills add hardikpandya/stop-slop
git clone https://github.com/mackswendhell/humanizer-pt-br.git .claude/skills/humanizer-pt-br

Step by step

1. One-time setup: feed the foundation skill with the product context: ` Here's my product context for product-marketing: [positioning, audience, differentiator] ` Every other skill consults that before executing any task.
2. Run an SEO audit on [your site]: a technical and on-page baseline before producing new content.
3. Write the article: ` Write a blog post targeting the keyword [keyword] for [audience] ` The copywriting skill consults product-marketing first, so the article comes out aligned with the real positioning, with CTAs coherent with the funnel.
4. The conversion lens: Optimize this article's CTAs and structure for conversions (the cro skill).
5. The GEO layer: Optimize this page to be cited by AI search (ChatGPT, Perplexity, AI Overviews) (the ai-seo skill): it reformats passages for LLM citability.
6. Text quality: the Stop Slop score (≥35/50) + Humanize este texto (PT-BR).
7. Ambiguous decisions about topic or angle: Ask the marketing council: [question]: a simulated panel of specialists (growth, brand, product, sales) debates before you invest in the article.

The strongest combination in practice

Stack 1 as the base + Stack 2's research pipeline on the pillar articles alone. The skills coexist with no conflict in the same .claude/skills/ folder, and Stack 2's information gain is the factor that most separates who ranks from who does not in competitive niches.

Golden rules (they hold for all 3 stacks)

1. Humanize before the technical validation: the rewrite can alter headings and break the seo-check.
2. Never publish a statistic without running the factcheck: an invented statistic destroys E-E-A-T.
3. A score is a gate, not a suggestion: an article below 90 (analyze) or 35/50 (Stop Slop) goes back for revision.
4. Monitor after publication: ranking gets built in the cycle: publish → measure (/blog google) → update decaying content (/blog audit).

Source: the Skills Hub catalog (index-skills-02.html), 44 skills and tools evaluated, 294 commands documented.