This skill turns a raw draft into an editor workflow. Instead of treating humanizer like a generic writing tool, the agent uses it to inspect a piece of prose for specific AI-writing fingerprints, such as inflated significance, vague attribution, promotional filler, rule-of-three phrasing, em dash overuse, chatbot artifacts, and mechanically uniform sentence rhythm. The agent can score a draft, surface the highest-risk passages, compare revisions, and suggest concrete edits that make the writing sound more natural without changing the author’s intent.

Invoke this when you already have text and the problem is credibility, tone, or authenticity. It is especially useful for blog drafts, landing-page copy, LinkedIn posts, emails, internal docs, and ghostwritten content that feels polished in the wrong way. In those cases, using the product normally as a standalone CLI only gives you raw analysis. The skill framing matters because the agent can decide when to scan, what to fix first, whether to ignore code blocks, and how to turn pattern findings into a practical revision pass.

The scope boundary is clear: this is not a general writing assistant, detector marketplace listing, or content generator. The skill is specifically about identifying and removing AI-style artifacts from existing text. Integration points are simple and verifiable: local files, Markdown docs, repo-wide scans for documentation folders, and CI checks that fail when a draft crosses a chosen score threshold.