The Anti-Slop Checker

The reason most AI-assisted writing fails isn't the grammar — it's the feel. Readers can't articulate why a paragraph "sounds AI" but they sense it within seconds, and that feeling kills trust. Marketing copy stops converting. Client deliverables get redrafted. LinkedIn posts get scrolled past.

This skill is a forensics tool, not a grammar checker. It scans for three layers at once: phrase-level tells (the hollow intensifiers, filler transitions, and beige adjectives that AI overuses because they're statistically safe), structural tells (uniform sentence length, perfect parallel structure, the lack of fragments or questions that a real writer would use), and register mismatches (where the formality doesn't fit the context).

It's context-aware. Academic writing isn't penalised for using "epistemological" — that's domain-appropriate vocabulary, not an AI tell. A legal brief is allowed to say "pursuant to". A SaaS landing page, on the other hand, gets the most aggressive pass because that register has the most AI contamination. The skill reads what you wrote, figures out what kind of writing it should be, then judges it against the right standard.

The output isn't just a score. It's a flagged table of every specific phrase with the line number, why it triggers the AI feel, and a rewritten version that preserves your meaning but reads like a person wrote it. Plus a "personal slop fingerprint" — the patterns you over-use — so you can catch them before they appear next time.

It's deliberately direct, sometimes blunt. The goal isn't to make text casual; it's to make it human. Those are different.