stop-sloplisted

# Stop Slop Eliminate predictable AI writing patterns from prose. ## Scope This skill targets stylometric tells: the phrases, structures, and rhythm that make prose read as machine-written. It works on what human readers notice, and on what style-based detectors (the GPTZero family) score. It does not touch model-level marking. Current models embed a watermark in their token choices, and C2PA provenance metadata in the files they generate. De-slopped prose is still marked prose. Never treat a stop-slop pass as removing provenance, and never describe it that way to anyone. See [references/detection.md](references/detection.md) for how the two layers differ and why the distinction matters when someone runs a detector on your work. ## Core Rules Slop is the statistical center of the distribution. A model samples from its highest-scoring candidates, and its highest-scoring candidates are the most predictable phrasings available. Every rule below pushes away from that center. When a rule feels arbitrary, ask the question underneath it: could a reader guess this word before reading it? 1. **Cut filler phrases.** Remove throat-clearing openers, emphasis crutches, and all adverbs. See [references/phrases.md](references/phrases.md). 2. **Break formulaic structures.** Avoid binary contrasts, negative listings, dramatic fragmentation, rhetorical setups, false agency. See [references/structures.md](references/structures.md). 3. **Use active voice.** Every sentence needs a human