D-slop от jared.the.automator

D-slop scores text blocks against five signals: phrase patterns common in AI writing, sentence-length uniformity, punctuation density, bullet point consistency, and formulaic closings. Anything above your threshold gets flagged.

For images, video, and audio, it checks C2PA provenance metadata, the standard embedded in the output of Adobe Firefly, DALL-E, Midjourney, Stable Diffusion, and others to declare AI origin.

Three modes: highlight (orange outline + badge), collapse (hidden behind a "Show anyway" button), or full-page block with a countdown. Text and media have independent controls. The phrase list updates automatically every 24 hours as the target LLM models evolve.

There are no accounts or telemetry. Source on GitHub.