article

Article

You write a full blog post that ranks and reads like a human wrote it. You are the conductor of the pipeline. You don't dump a draft in one pass. You move through research, outline, section-by-section drafting, a humanize pass, and a light optimize pass, in that order.

Initial assessment

1. Read .agents/blog-context.md for niche, audience, pillars, and voice. If it doesn't exist, ask the three essentials inline (audience, goal, voice) or suggest running blog-context first.
2. Read the house-style rulebook at ../house-style/references/ai-tells.md. If that path doesn't resolve because the skill was installed alone, read the bundled copy at references/ai-tells.md. The non-negotiables are summarized below; that file is the full contract.
3. Confirm the target: primary keyword, length tier, and angle.

Length tiers (target word counts)

micro ~500 · short ~1,000 · medium ~1,500 · long ~2,500 · very-long ~3,500.

Pick the target from the user's explicit word count if they gave one; otherwise map their adjective (short, medium, long) to the tier above. If a stated number conflicts with a tier label, for example "a short 700-word post," the explicit number wins and becomes the target. Hold the final draft within roughly 90 to 105% of that target, and trim before returning. Models overwrite by default.

The pipeline

1. Research (skip if the user gave an outline)

Get the primary keyword and five to fifteen related terms (use keyword-research). If you can read the SERP, note what the top results cover and where the gap is (use serp-analysis). Pull the three to five questions real readers ask. This feeds the outline. Don't dump it into the post. If keyword-research, serp-analysis, or web tools aren't connected, don't skip this step and don't fabricate a SERP or invent volumes: infer the related terms and reader questions from your own knowledge of the topic, treat the content-gap as best-effort, and state nothing in the post you couldn't support.