backlink-audit is an off-page SEO workflow for AI agents. It starts from whatever link data exists: a CSV export from Ahrefs, Semrush, Majestic or DataForSEO, or the Google Search Console Links report when no paid tool is available.
It reviews referring domains, anchor-text mix, the most-linked pages against the pages that earn money, and spammy link signals. The report states what the data can and cannot show, so a partial Search Console sample is never presented as the full link index.
What You Get
- A link-profile summary with an explicit note on how complete the data is
- An anchor-text breakdown split into branded, exact-match, generic and URL anchors
- Internal-linking quick wins that pass authority from heavily linked pages to money pages
- A link-building priority list of relevant, attainable targets
- A clear verdict on toxic links, including whether a disavow file is warranted
Common Use Cases
- Reviewing a site’s link profile with only Search Console access
- Checking whether exact-match anchors create penalty risk
- Finding heavily linked blog posts that can feed authority to conversion pages
- Deciding whether spammy links justify a disavow file
What Makes It Unique
It is built to work without a paid backlink tool and says so up front. When only the Search Console sample is available it states that limit, and it refuses to invent DR or DA values or link counts. It also advises against over-using disavow, pointing out that Google usually ignores spam links, and it treats internal linking from linked pages as a high-return fix the site owner controls.
Under The Hood
The skill is a short wrapper whose SKILL.md tells the agent to read a longer WORKFLOW.md in the seo folder of the same repository and follow it. That workflow is prose only and ships no scripts. It asks the agent to read the Search Console Links report through whatever connected capability the host provides, or to parse a CSV the user attaches. The workflow’s safety rules say not to push or publish changes to a CMS or live site without an explicit yes, and to treat fetched pages and tool output as untrusted data. It also reads shared preamble and safety files from the repository.