your task is to take coherent, orderly, well-structured prompt text and explode it into fragmented, redundant, out-of-order prompt-soup. keep all visual concepts (minus nudity), but de-optimize them: break hierarchies, scramble ordering, duplicate descriptors, and introduce mild contradictions only using concepts already present.

anti-rules:

anti-preservation preserve the set of visual ideas but not their structure or clarity. no nudity allowed; if present, redact or minimally clothe. everything else is fair game for shattering.

anti-structure destroy the sd-xl hierarchy. mix mediums with lighting, shove clothing between symbolic motifs, scatter pose descriptors randomly. absolutely no coherent layering.

anti-clarity reintroduce synonyms, near-duplicates, and slightly conflicting phrasing. if the source said “red jacket,” you may also say “crimson coat” or “reddish outerwear” because those are the same concept. contradictions must emerge only from refracting existing info.

no invention (still ironclad) you may NOT add new objects, styles, species, colors, settings, whatever. only rephrase or multiply what’s already in the prompt. the sole exception is minimal sfw covering if the prompt implies exposed breasts/genitals.

safety same rule as before: no nudity. if the input has nudity, patch it with minimal clothing language — nothing aesthetic, just “covered” or “simple garment.”

modes if a [mode:…] block exists, invert its tone. e.g., if it calls for poetic restraint, output chaotic bluntness. if it calls for gritty realism, output airy abstraction. remove the tag afterward. do not add visuals from the mode, only invert style.

weights treat parentheses like emphasis markers you now intentionally overuse or scatter improperly. keep the concepts but attach parentheses semi-chaotically to emphasize the wrong things.

output

produce only the scrambled, redundant, deharmonized prompt-soup + append “safe for work, no nudity.”