| name | compress |
| description | Compresses natural language files (markdown, text) into caveman-speak to reduce input token usage ~46%. Backs up original before overwriting. Only touches prose — code blocks stay exact. Trigger: /caveman:compress <filepath>, or user asks to compress a memory/doc file.
|
Compress target file to caveman prose. Backup first. Code blocks exact. Retry up to 2x on error.
Trigger
/caveman:compress <filepath> or user asks to "compress this file", "shrink this doc", "make this memory smaller".
Allowed file types
.md, .txt, .typ, .typst, .tex, extensionless files only.
Never touch: .py, .js, .ts, .json, .yaml, .yml, .env, .toml, .sh, or any code file. Never re-compress .original.md files.
Process
- Read target file
- Create backup:
<filename>.original.md (same dir) — skip if already exists
- Compress prose sections (see rules below)
- Write compressed content back to original file path
- Report: original token count, compressed token count, % reduction
Compression rules
Remove: articles (a/an/the), filler (just, really, basically, actually, simply, in order to → to), pleasantries (please, feel free to, don't hesitate), hedging (you might want to, consider, perhaps), "you should" constructions.
Preserve exactly: code blocks (...), inline code (...), URLs, file paths, shell commands, technical terms, proper nouns, dates, version numbers, env vars, error strings.
Preserve structure: markdown headings, bullet hierarchies, numbered lists, tables, YAML frontmatter.
Compress text: shorter synonyms, sentence fragments OK, merge redundant bullets into one, keep single representative example when multiple exist.
Example:
- Before: "You should always run your tests before pushing your changes to make sure everything is working correctly."
- After: "Run tests before push."
Validation
After compression:
- All code blocks present and unmodified
- No URLs dropped
- File is valid markdown (headings intact, no broken lists)
- Token count reduced (if not, warn user)
If validation fails: retry compression up to 2 times. If still failing: report error, restore original, do not overwrite.
Output
Compressed <filepath>
Original: ~N tokens → Compressed: ~N tokens (X% reduction)
Backup: <filepath>.original.md
