When you already have the outline and just need it executed, humanized, and verified.
Here is my outline: [paste your outline] Write a [word count]-word blog post following this outline exactly. Target keyword: "[keyword]". Preserve these phrases: [phrase 1], [phrase 2], [phrase 3]. Humanize with Walter in balanced mode. Run detection. Show the keyword report.

Your outline becomes a finished article. Same pipeline.
Output: Article matching your outline + scores. Time: ~2 min.
Inputs to have ready before you start: paste your outline, word count, keyword, phrase 1, phrase 2 and phrase 3. Each bracketed field in the template becomes a box in the form above, so you can fill them once and copy the finished prompt.
Blog prompts work best when you give Claude something it cannot invent: your angle, a real example, a number you can stand behind, and the exact phrase you want to rank for. The prompt handles structure and the Walter steps; your inputs decide whether the post says anything new.
Before you publish, read the draft once as a reader, not an editor. Check every claim, add the sources you used, and cut any paragraph that repeats the one before it. Google's guidance is about whether a page helps the person who landed on it, not how the words were produced, so the review step matters more than the detection score.
Treat the detection score as a signal, not a verdict. A high score is a reason to reread the sentences Walter flags; a low score means you can move on to the fact check, not that the piece is finished. Your own read of the final draft is the last step.
Skills are project-level instructions that pre-load Walter, mode, and keyword rules, so you can drop a one-line brief and skip the boilerplate every time.
Browse the Skills libraryWalter installs as an MCP connector in about 30 seconds. Free tier covers your first articles, no card.