Agentic SEO Prompts — Development skill for Claude Code

Agentic SEO — Agent A workflow prompts

Companion repo for the article **["What is agentic SEO and how do you get started?"](https://ahrefs.com/blog/agentic-seo/)**. Each section is a copy/paste prompt that recreates one of the eight workflows in the post (plus a Bonus).

Every prompt:

• Names the Agent A skill it mirrors, so users running Agent A know which prebuilt skill is doing the work behind the scenes.
• Calls out the 1–2 tricks that make the workflow actually work — the things a generic prompt would miss.
• Ends with a strict output schema (Markdown / CSV) so the agent returns something you can paste into a spreadsheet, ticket, or doc — not a wall of text.

Replace anything in **[brackets]** with your own values.

---

1. Site Audit Triage (with optional PR fix)

*Maps to workflow 1 in the article. Agent A skills: `ahrefs-site-audit` + `ahrefs-site-audit-fix`.*

Run an Ahrefs Site Audit for **[your domain]**, triage the issue list into a ranked fix queue for this sprint, and — optionally — open a pull request with the fix for the top issue.

**Inputs**

• Domain / crawl scope: [your domain]
• Country / language focus: [country + language, or "global"]
• Sprint capacity: [hours or person-days available this sprint]
• (Optional) Repo / location to change: [GitHub repo URL + branch, or local path] — only needed if you want a PR

**Part A — Triage**

1. Pull the latest crawl's health score + full issue list. Note total pages crawled.
2. Score each issue by expected impact on (a) organic traffic and (b) crawl efficiency.
3. Assign a fix strategy to each issue (the part that turns a flat list into a real queue):
   • template — fix shared layout/template; one edit fixes many pages
   • per-page — fix needs individual page context
   • config — fix in sitemap, robots.txt, or server config
   • investigate — present data; don't auto-fix; needs human call
4. Estimate effort. Drop duplicates / very low-signal items. Consolid