Cut communication token usage 75%

Cut token usage ~75% by dropping filler while preserving technical accuracy

When technical writing must fit tight token budgets without sacrificing precision (e.g., API docs, complex prompts, reference material).

• · full-form technical text or prompt

• · ultra-compressed version: no articles, filler, pleasantries
• · token-savings estimate
• · caveman-style lint warnings

• · stdlib Python tools (text compressor, token-savings estimator, caveman linter)
• · 3 reference docs: compression principles, technical communication patterns, when caveman backfires
• · cs-caveman-mode persona agent
• · /cs:caveman slash command

• · text is technical (caveman breaks for marketing/narrative copy)
• · accuracy preservation is non-negotiable
• · token count matters (cost or latency budget)

• · Over-compression on ambiguous domain language obscures meaning
• · Caveman linter flags legitimate technical jargon as filler
• · Drop-articles pass makes some sentences grammatically incorrect

• · 75% token reduction documented
• · stdlib Python tools (no external deps)
• · 3 deep references + when-to-use guardrails
• · Derived from Matt Pocock MIT caveman skill (voice/rules preserved)
• · cs-caveman-mode persona agent + /cs:caveman command