seo-blog-writer

What it does

End-to-end pipeline for shipping a single long-tail blog post: topic -> research -> draft -> scrub -> AI-SEO audit -> publish. The writing is platform-agnostic, producing a publish-ready bundle (clean HTML, slug, meta, JSON-LD blocks, feature-image alt) with a pluggable publish step (Ghost Admin API, WordPress REST, or static-site file output). It is designed for SEO and AI-citation extractability (FAQ blocks, BreadcrumbList + FAQPage + HowTo schema, query-phrased headings). It takes one required argument: the topic, with optional flags to control the publish target and state. The default state is draft; the default target is static (writes a self-contained HTML file + metadata). The skill supports publishing via Ghost, WordPress, or static-site outputs, and can schedule future publishes with an ISO timestamp. The preflight checks ensure Python availability and writable directory, with optional integration to an ai-seo MCP for programmatic citation-worthiness and schema scoring. It is designed for Claude Code and uses a platform-agnostic core with a pluggable publish step.

How it works

• Step 0: Parse the topic argument and classify the long-tail query into a template (e.g., troubleshooting, integration, explainer, etc.). Angle modifications via --angle can append to the topic.
• Step 1: Research the topic via WebSearch on the topic (and with current year). Seed search results to extract: (a) word-count distribution from top results, (b) People Also Ask boxes to capture verbatim for FAQ, (c) any featured snippet format to emulate lead paragraph.
• Step 1c–1e: Deep fetch 2–4 URLs (vendor docs, GitHub issues/changelogs, forums, existing top posts). Save research artifacts and outbound interlinks (tmp/blog-drafts/<slug> files) for Step 3 integration.
• Step 1d: Answer five gating questions before drafting: exact intent, direct answer, canonical primary source, notable gotcha, and 3–6 FAQ follow-ups. If uncertain, continue research or ask for sources.
• Step 2: Choose format and length band based on template mapping; target word count within 600–1500 words for most formats, using SERP signals to adjust within the band.
• Step 3: Draft the post in HTML with allowed tags, following a TL;DR block as the first paragraph, lead paragraph, and H2s as questions or allowed labels. Apply link policy for internal vs. outbound with appropriate rel attributes. Maintain an impersonal voice and precise details (version numbers, error messages, etc.). Ensure structured sections for FAQ and other schema where applicable.
• Step 0–7: The pipeline proceeds up to platform credentials, with the publish step handled by adapters for Ghost, WordPress REST, or static files; default is draft unless --publish or --publish-at is provided.

When to use it

Use when you want to generate a publish-ready blog post from a single long-tail query, with an emphasis on AI-citation and structured data (FAQ, BreadcrumbList, HowTo JSON-LD). Ideal for indie hackers, founders, and content marketers seeking a citable post rather than paraphrased docs. Trigger phrases include: 'write a blog post on X', 'draft an article about X', 'publish a post on X to Ghost / WordPress / the static site', or any request to ship editorial content for a long-tail query.

What it can touch

The skill touches the platform adapters for publishing: Ghost Admin API, WordPress REST, or static-site file output. It can write a self-contained HTML file and a metadata.json when targeting static-site output. The default target is static, and credentials are only required at the publishing step (Step 8) for the specific adapter.

Caveats

• License: MIT-0 as stated in frontmatter. - Requires: Designed for Claude Code. - Platform credentials are only needed at the publish step; the writing pipeline itself runs without credentials. - The open-source repository indicates a dependency on Python for rasterizer, scrubber, and schema builder. - The AI MCP integration is optional; if not connected, manual audit proceeds.