Authenticity Check — AI skill for Claude Code

authenticity-check

      

A standalone, pure-prompt skill that scores how authentically a piece of text reads as the work of a real human author, and flags the specific spans that read as AI-generated, AI-templated, or generically derivative. It also scans supplied text for suspicious Unicode provenance carriers and reports them in a separate evidence channel. It is just instructions: `SKILL.md` plus a few reference files. No scripts, no dependencies, no network access. Its tools are read-only by design.

It is the evaluative counterpart to the `humanizer` skill. This one diagnoses. It does not rewrite.

Where this comes from

This skill is the diagnostic half of a pair. Its detection criteria descend from the voice-preservation logic that powers [Scriveno](https://github.com/hannsxpeter/scriveno) (formerly Scriven), an AI-native longform writing, publishing, and translation pipeline whose core promise is that drafted prose should sound like the writer, not like AI. [`humanizer`](https://github.com/hannsxpeter/humanizer) lifted the de-slop, restraint, and voice-matching layer of that pipeline into a standalone rewrite skill. `authenticity-check` is the read-only counterpart: it applies the same catalog and the same restraint to *diagnose* rather than transform, and it vendors humanizer's criteria so the two agree on what a tell is. If you want the rewrite, use humanizer. If you want the whole writing pipeline, see Scriveno (npm package: `scriveno-cli`).