content-writer is a three-job content workflow for AI agents: a new blog post, a new landing page, or an improvement of an existing page. It infers the job from the request, gathers keyword, audience and brand context, researches the search results when web access exists and shows an outline for approval before writing.
Blog posts must clear a fixed bar: a hook-driven title, a table of contents, at least 1000 words, a featured image with three or more inline images, an FAQ and applicable JSON-LD.
What You Get
- A full markdown post or page with a clean heading hierarchy and table of contents
- SEO metadata: title tag, meta description and URL slug
- JSON-LD structured data matched to the content
- Image roles, placements, alt text and generation prompts
- An internal linking plan and a publishing checklist
- For rewrites, a change summary tied to specific guideline violations
Common Use Cases
- Drafting a how-to blog post for a target keyword
- Writing a service, product, location or comparison landing page
- Rewriting a thin or stale page that is not ranking
- Turning a planned topic from a content calendar into a finished article
What Makes It Unique
It sets a high floor for blog posts: hook-driven titles, at least 1000 words, a featured image plus three inline images that explain something, and a Last Click test asking whether the reader would need to search again. It redirects intent mismatches (a transactional keyword gets a landing page, not a post) and treats compliance-sensitive topics as needing official sources. If no image generation capability exists it hands over prompts and labels the missing assets without claiming generation succeeded.
Under The Hood
The skill is a short wrapper whose SKILL.md tells the agent to read a longer WORKFLOW.md in the seo folder of the same repository and follow it. The workflow contains an embedded shell command that runs find over the plugin and skill folders in the home directory to locate the content-writing reference file. It also has the agent search the web for the target keyword and read pages when web access exists, and call an image generation capability and save outputs to a location the user names. It never pushes content to a CMS or live site without an explicit yes, and treats fetched pages as untrusted.