X Article Agent Skill

Format a post for an X Article

X Articles only support two heading levels ("heading" and "subheading") plus bold text, and images must be uploaded one at a time. This skill produces a clean markdown file and a folder of ordered images so the user can copy/paste text and drag images in order.

The user will provide a post — a slug, a filename, or a path under contents/blog/ or contents/newsletter/: $ARGUMENTS

Step 1: Locate the source

Find the source file under contents/blog/ or contents/newsletter/ (extension may be .md or .mdx). If the argument is ambiguous or matches multiple files, ask which one. Note whether it came from blog or newsletter — the CTA in Step 4 only applies to newsletters.

Read the full file.

Step 2: Set up the output

Derive {slug} from the source filename (without extension). Create the output folder x-articles/{slug}/ (relative to repo root) and an images/ subfolder inside it.

Step 3: Write x-articles/{slug}/article.md

Transform the source body into the article file:

• Remove the frontmatter (everything between the opening and closing ---).
• Remove MDX import lines and any other JSX/component noise that won't render as plain text (e.g. <NewsletterForm />, <ProductComparisonTable .../>, wrapping <div>s). If a component can't be represented as text, drop it. Images inside components are still handled in Step 4.
• Title → H1. Take the title from the frontmatter and put it as a single # Title at the very top of the file.
• Remap headings (because X collapses everything to two levels):
• ## → #
• ### → ##
• #### Heading → Heading (bold paragraph, heading text only)
• Leave #####+ as bold too.
• Keep all body text, links, lists, blockquotes, code blocks, and footnotes verbatim. X has no relative-link context, so prefix any relative links (e.g. /blog/..., /feature-flags) with https://posthog.com.
• Replace each image with a numbered marker. Walk the body top to bottom and replace every image (see Step 4 for what counts as an image) with a line like:

   [IMAGE 1: alt text]

Numbered in document order starting at 1. This tells the user which downloaded image goes where.

Step 4: Download the images

Collect every image URL in document order. Images appear as:

• Markdown: !alt
• Components: imageLight="https://res.cloudinary.com/..." (and imageDark= — prefer imageLight; skip the dark duplicate)
• Raw <img src="...">

Cloudinary URLs may contain transformation segments between /upload/ and the filename (e.g. /upload/q_auto,f_auto/file.jpg or /upload/w_1600,c_limit,q_auto,f_auto/file.jpg). Download the URL as-is — the transformations are harmless.

Do not download the featuredImage from frontmatter unless it also appears in the body.

For each image in order, save it to x-articles/{slug}/images/{n}.{ext} where {n} matches the [IMAGE n: ...] marker and {ext} is the original extension from the URL (jpg, png, gif, webp). Download with:

curl -sSL -o "x-articles/{slug}/images/{n}.{ext}" "{url}"

Run the downloads and verify each file is non-empty (a failed Cloudinary fetch returns a tiny error body). Report any that failed.

Step 5: Newsletter CTA

If — and only if — the source was a newsletter, append this to the very bottom of article.md:

> Subscribe to [build mode](https://newsletter.posthog.com/subscribe), our newsletter, to get more writing like this.

Step 6: Report

Tell the user:

• The output path x-articles/{slug}/article.md
• How many images were downloaded (and where), and flag any failures
• A reminder that [IMAGE n: ...] markers in the text show where each numbered image goes