“Site Audit Skills”
This set of skills is for running a technical SEO audit: diagnosing crawl findings, grouping them and turning them into action items.
Skills are guides the assistant follows when using the Ranki MCP server. Site Audit is read-only and doesn’t use API credits.
For the full audit, ask for ranki-site-audit-full. Example: “Run the full technical audit for the Online Store project.” Include the project name exactly as it appears in Site Audit. If you don’t, the assistant shows you the projects in your account and waits for you to pick one.
In Claude, you can pick the skill by name from the prompt list. In Cursor, just ask for it in the chat.
What each one does
ranki-site-audit-full
Orchestrates the full audit. With a single request, it runs, in this order:
ranki-site-audit-roadmap(and that skill callsdiagnoseandtasksfor each check and creates the rows).ranki-site-audit-architecture, thentasksand the rows for those cases.ranki-site-audit-group, once the initial tasks exist. It proposes the groups and waits for your approval.
Use it when you want the full technical audit. For a single issue, just the tree or just the issues, call the skill for that step manually.
ranki-site-audit-roadmap
It’s the shortcut for the issues audit. With a single request, it goes through the crawl checks and, for each one, runs diagnose and then tasks, and creates one page in SEO Tech Roadmap per actionable case. One check per pass.
That’s where diagnose and tasks run on their own. group isn’t part of that pass. Group in a follow-up message, once those tasks exist. Requires Notion.
For a single issue, or a few, call diagnose and tasks manually. The roadmap is designed for the full issues crawl, not for architecture or grouping.
ranki-site-audit-diagnose
Explains a crawl issue: 404s, broken links, titles, H1s, meta descriptions, redirects, canonicals or noindex. It splits the cases into Problem, Cause and Solution, with the URL table.
Use it manually when you want one or more specific checks. If you ask for the full roadmap, you don’t need to name it: ranki-site-audit-roadmap runs it on every check.
If the request is about the URL tree rather than a check, the assistant switches to ranki-site-audit-architecture.
ranki-site-audit-tasks
Turns the diagnosis already in the thread into tasks: title, problem, cause, solution and priority.
Runs after diagnose. If you ask for tasks and there’s no diagnosis yet, the assistant diagnoses first and then builds the tasks.
Use it manually after a standalone diagnosis or after architecture. Inside the roadmap you don’t need to name it: the roadmap runs it on its own, on every check, right after diagnose.
By default, these tasks are created as rows in Notion. You can ask for another destination: other software connected to the assistant, or a text file. Say so in the same message. Example: “…and save them to a text file” or “…and create them in Linear.”
ranki-site-audit-group
Reads the cards already created in SEO Tech Roadmap and proposes merging the ones that can be closed with a single change. It proposes and waits for your approval. If you approve only some, it merges only those.
Do the grouping after you’ve created the initial tasks. Before that, there are no cards to merge. This skill doesn’t diagnose or create tasks: it reorganizes what’s already in the roadmap. Requires Notion connected in the same assistant.
ranki-site-audit-architecture
Diagnoses the URL tree: languages, folders, depth, duplicate titles and pages cloned across countries. 404s, redirects, canonicals and noindex stay in diagnose; architecture doesn’t cover them.
It doesn’t create tasks or write to Notion on its own. Afterward, call tasks manually if you want action items. If you want the roadmap rows, ask for them in a separate message. Grouping comes once those tasks exist.
Usage examples
1. Full technical audit
Request:
“Run the full technical audit for the Online Store project.”
Runs on its own ranki-site-audit-full. That skill calls roadmap (which brings in diagnose and tasks), then architecture with its tasks and rows, and finally group. Don’t name the other skills.
Grouping is proposed once the initial tasks exist. The assistant waits for your approval before merging.
If you don’t specify a destination, tasks go to Notion. In the same request, you can ask for other software or a text file: “…and save the tasks to a text file.”
2. Diagnose, create tasks and group one or more issues
Request:
“Diagnose the 404s and duplicate titles in the Online Store project and turn them into roadmap tasks.”
Here you call them manually: diagnose on each issue you named and tasks on that diagnosis. The full roadmap isn’t involved; only those checks, and the tasks end up as rows.
Once those tasks exist, a second message:
“Group the 404 and duplicate title tasks.”
group doesn’t run inside diagnose or tasks. Do it after you create the initial tasks.
3. Architecture, tasks, grouping and roadmap
Request:
“Audit the URL architecture of the Online Store project.”
Runs on its own ranki-site-audit-architecture. Check diagnosis isn’t involved.
Next message, manually:
“Turn those cases into tasks and create them in the roadmap.”
That’s where tasks and the Notion writing come in. architecture doesn’t trigger them on its own.
Once those tasks exist:
“Group the architecture cards.”
Where tasks end up
Unless you say otherwise, the roadmap creates them in SEO Tech Roadmap (Notion). The assistant needs the Notion connector in the same client, in addition to the Ranki MCP.
You can ask for another destination in the same message:
- other software the assistant has connected (a board, a task manager, a spreadsheet);
- a text file, with one task per case (title, problem, cause, solution and priority).
Grouping cards with ranki-site-audit-group reads and merges Notion rows. If tasks go to other software or a file, ask for grouping on that same output: the assistant proposes what to merge and waits for your approval, without touching Notion.
The parameters for the data tools are in Site Audit.
What they don’t do
- They don’t start a new crawl or edit the site.
- They don’t spend keyword research credits.
- They don’t make up URLs, status codes or titles that aren’t in the crawl.