A Codex skill for losslessly compressing AI image and video generation prompts into dense, mostly Simplified-Chinese hybrid prompts.
It preserves generation-relevant details such as reference-image rules, subject consistency, timelines, causal action, camera language, VFX/SFX, spatial relationships, exact numbers, positive constraints, and negative prompts. When a prompt cannot fit a requested MAX_CHARS limit without semantic loss, the skill emits an explicit warning instead of silently dropping instructions.
Install this repository with the Codex skill installer, or copy the repository folder into ~/.codex/skills/lossless-prompt-packer.
Invoke $lossless-prompt-packer, provide an image/video-generation prompt, and optionally specify a hard limit such as MAX_CHARS=5000.
The included scripts/count_chars.py helper counts Unicode code points and can validate a hard limit.