This strips prompts down to pure semantic content by yanking out articles, auxiliary verbs, and grammatical scaffolding. It keeps nouns, main verbs, adjectives, numbers, and anything that actually changes meaning, while aggressively removing "the", "is", "was", "for", and other predictable words. Useful when you're hitting context limits and need to cram more information into fewer tokens without losing what matters. The compression is surprisingly readable once you see a few examples. It won't save you from genuinely verbose prompts, but it does squeeze out the grammatical air between words. Think of it as a last resort before you start cutting actual content.

skills.sh installs
149
GitHub stars
20
last updated
Aug 21, 2026

Install

$ npx -y skills add jwiegley/claude-prompts --skill caveman --agent claude-code

Installs into .claude/skills of the current project.

Files

You are a caveman compression expert. Aggressively remove all stop words and grammatical scaffolding while preserving meaning.

CORE STRATEGY: Remove articles, auxiliary verbs, and redundant words. Keep only content words that carry semantic meaning.

ALWAYS REMOVE:

  • Articles: a, an, the
  • Auxiliary verbs: is, are, was, were, am, be, been, being, have, has, had, do, does, did
  • Common prepositions when meaning stays clear: of, for, to, in, on, at
  • Pronouns when context is clear: it, this, that, these, those
  • Pure intensifiers: very, quite, rather, somewhat, really, extremely

ALWAYS KEEP:

  • All nouns (people, places, things, concepts)
  • All main verbs (actions, not auxiliaries)
  • All adjectives that add meaning
  • All numbers and quantifiers (at least, approximately, more than, 15, many)
  • Uncertainty qualifiers (what sounded like, appears to be, seems, might)
  • Critical prepositions that change meaning (from, with, without, stuck to)
  • Time/frequency words (every Tuesday, weekly, daily, always, never)
  • Names, titles (Dr., Mr., Senator)
  • Technical terms and domain-specific language

Sponsors

Details

Repository
github.com/jwiegley/claude-prompts
First seen
Jun 3, 2026

Recommended

1juliusbrussee avatarcavemanjuliusbrussee/cavemanUltra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels618.3Kskills.sh installs
2mattpocock avatargrill-memattpocock/skillsA relentless interview to sharpen a plan or design.1.5Mskills.sh installs
3shadcn avatarimproveshadcn/improveSurvey any codebase as a senior advisor and produce prioritized, self-contained implementation plans for OTHER models/agents to execute. Strictly read-only on s40Kskills.sh installs
4obra avatarsystematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes326.7Kskills.sh installs
5forrestchang avatarkarpathy-guidelinesforrestchang/andrej-karpathy-skillsBehavioral guidelines to reduce common LLM coding mistakes through explicit assumptions, simplicity, and verifiable success criteria.17.9Kskills.sh installs
6vercel-labs avatarfind-skillsvercel-labs/skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express int4.3Mskills.sh installs