Skip to content

Agent recipes for ScaleBop SEO

Once your agent is connected to ScaleBop SEO, you can ask it anything about your sites. These three routines are the ones worth running regularly. Each one lists the prompt to use and what the agent does with it, so you can check its work.

To run them without retyping the prompt, install the Claude Skill at the bottom of this page.

Ask: “Run my weekly SEO check.”

The agent:

  1. Lists your projects and, for each one, gets the overview: audit score, open issues by severity, and how old the audit is.
  2. Gets the audit diff: what’s new, fixed or worse since the previous audit.
  3. Compares Search Console clicks and impressions for the last 7 days with the 7 days before.
  4. Lists page changes from the past week, including anything the ScaleBop agent applied.
  5. Reports, per site: what got worse, what got fixed, the traffic change, and the one thing to do next.

A weekly check is only as fresh as the latest audit. If the agent says the audit is more than a week old, run one in the app first.

Ask: “Why did traffic to /pricing drop?” (or the whole site, or a search query)

The agent:

  1. Gets Search Console trends for that page or query over 90 days, and finds the days the drop started.
  2. For the days around the drop, looks for causes:
    • site events: your notes in ScaleBop and confirmed Google algorithm updates;
    • page changes: title, description, heading, canonical or content changes on that page;
    • the audit diff: new issues on that page, such as noindex or a changed canonical.
  3. Reports the most likely cause with its evidence, and says which other causes it ruled out.

A drop that lines up with an algorithm update is a candidate, not proof. If nothing lines up, the agent should say so rather than guess, and suggest what to check next, such as a competitor or a change made outside ScaleBop to a page that isn’t monitored.

Ask: “Fix the top 5 issues on my site.” Run this in Claude Code or Cursor, opened in your site’s code.

The agent:

  1. Lists open audit issues and picks five, starting with critical and high severity, and pages that affect many URLs.
  2. For each one, gets the fix written for your stack (Next.js, Astro, Lovable and others) and applies it to your code. For a site builder or CMS such as WordPress, Shopify or Webflow, it gives you the steps instead.
  3. After you deploy, rechecks each fixed page and reports what’s confirmed fixed, what’s still open, and what can only be confirmed by the next full audit (duplicate titles, orphan pages, broken links).

Review the agent’s changes before you deploy them, as you would any change to your site.

A Claude Skill teaches Claude Code these three routines, so “run my weekly SEO check” works in any session.

  1. Create the folder ~/.claude/skills/scalebop-seo/ (or .claude/skills/scalebop-seo/ inside one project).
  2. Save the text below in that folder as SKILL.md.
  3. Restart Claude Code. The skill loads when you ask for one of the routines.
---
name: scalebop-seo
description: Weekly SEO check, why-did-traffic-drop diagnosis, and fix-the-top-5-issues routines over the ScaleBop SEO MCP server. Use when the user asks for an SEO check, why traffic or rankings dropped, or to fix SEO issues on their site.
---
# ScaleBop SEO routines
These use the `scalebop` MCP server's tools. Start every routine with
`list_projects`; tools take project ids like `prj_…`, never numbers.
Keep the provenance each tool returns. Say where a number came from and
how old it is, never mix a Search Console position with a scraped one,
and call an algorithm update a candidate cause, not proof.
## Weekly SEO check
For each project:
1. `get_project_overview`: score, open issues by severity, audit age. If
the audit is older than 8 days, say so and suggest running one.
2. `get_audit_diff`: new, fixed and regressed issues.
3. `get_search_console_trends` with `days: 7`: compare `totals` with
`previousTotals`.
4. `list_page_changes` with `days: 7`.
5. `list_competitors`: anything a competitor added since the check before.
Report per site: what got worse, what got fixed, the traffic change,
competitor moves worth answering, and one next step.
## Why did traffic drop?
1. `get_search_console_trends` with `days: 90`, filtered by `page` or
`query` if the user named one. Find the dates the drop began.
2. Around those dates (a week either side):
- `get_site_events` for notes and algorithm updates;
- `list_page_changes`, filtered by `url` for a single page;
- `get_audit_diff` for new issues on the affected pages.
3. Give the most likely cause with its evidence and list what was ruled
out. If nothing lines up, say so; don't invent a cause.
## Fix the top 5
1. `list_audit_issues`. Pick five: critical and high severity first, then
the issue types that affect the most pages.
2. For each, `get_fix` with the issue type and an affected `pageUrl`.
Apply code fixes in this repository; for a CMS or site builder, give
the user the steps.
3. Ask the user to deploy. Then call `request_recheck` for each fixed
page with its `findingType`, and report fixed, still open, and needs a
full audit.
Never use `run_agent` or `revert_agent_action` unless the user asks.