Caveman Token Optimizer
Install when you want Claude Code or Codex answers in ultra-compressed caveman-speak to slash output tokens without losing technical precision.
Overview
Caveman Token Optimizer is a journey-wide agent skill that compresses Claude Code and Codex replies into caveman-speak—usable whenever a solo builder needs shorter agent output before burning context on filler.
Install
npx skills add https://github.com/aradotso/trending-skills --skill caveman-token-optimizerWhat is this skill?
- Strips pleasantries, hedging, and filler articles while keeping code blocks and exact technical terms
- Reports ~65% average output token reduction with peaks up to ~87% in marketing copy
- Preserves normal git commits, PR text, and verbatim error quotes
- Install via npx skills add or Claude marketplace plugin caveman@caveman
- Explicit trigger phrases like enable caveman skill and less tokens please
- up to ~87% reduction cited for marketing examples
- strips pleasantries hedging and articles while keeping code blocks intact
Adoption & trust: 672 installs on skills.sh; 31 GitHub stars; 2/3 security scanners passed (skills.sh audits).
What problem does it solve?
Your coding agent burns thousands of output tokens on pleasantries and hedging even when you only need a fix or snippet.
Who is it for?
Solo builders running long Claude Code or Codex sessions who already understand terse answers and want measurable token savings.
Skip if: Teams shipping customer-facing docs, onboarding copy, or stakeholder emails that require polished professional tone without a separate rewrite pass.
When should I use this skill?
User asks to install caveman mode, enable caveman skill, reduce token usage caveman, or wants less verbose Claude Code replies.
What do I get? / Deliverables
After enabling caveman mode, replies stay technically accurate with roughly two-thirds fewer output tokens on average.
- Installed caveman skill or plugin
- Compressed caveman-style agent responses for subsequent turns
Recommended Skills
Journey fit
Useful at every journey phase - explore requirements and options before committing to a direction.
Where it fits
Run caveman mode while iterating API fixes so stack traces and patches arrive without paragraph-long preambles.
Keep test failure analysis terse during a pre-release regression pass with your agent.
Triage production logs with compressed remediation steps when every token counts in a long thread.
Draft internal analytics notes with the agent—then switch tone off before publishing outward-facing copy.
How it compares
Use as a response-style plugin, not as an MCP server or codebase analysis tool.
Common Questions / FAQ
Who is caveman-token-optimizer for?
Indie developers and small teams using Claude Code or Codex who want the same technical answers with far less chatty output token spend.
When should I use caveman-token-optimizer?
Enable it during build and ship debugging marathons, operate incident triage with your agent, or any session where verbose defaults waste context—trigger with install caveman mode or less tokens please.
Is caveman-token-optimizer safe to install?
It mainly changes how the model phrases answers; review the Security Audits panel on this Prism page before adding marketplace or npx skills from third-party repos.
SKILL.md
READMESKILL.md - Caveman Token Optimizer
# Caveman Token Optimizer > Skill by [ara.so](https://ara.so) — Daily 2026 Skills collection. A Claude Code skill and Codex plugin that makes AI agents respond in compressed caveman-speak — cutting **~65% of output tokens on average** (up to 87%) while keeping full technical accuracy. No pleasantries. No filler. Just answer. ## What It Does Caveman mode strips: - Pleasantries: "Sure, I'd be happy to help!" → gone - Hedging: "It might be worth considering" → gone - Articles (a, an, the) → gone - Verbose transitions → gone Caveman keeps: - All code blocks (written normally) - Technical terms (exact: `useMemo`, `polymorphism`, `middleware`) - Error messages (quoted exactly) - Git commits and PR descriptions (normal) **Same fix. 75% less word. Brain still big.** ## Install ### Claude Code (npx) ```bash npx skills add JuliusBrussee/caveman ``` ### Claude Code (plugin system) ```bash claude plugin marketplace add JuliusBrussee/caveman claude plugin install caveman@caveman ``` ### Codex 1. Clone the repo 2. Open Codex inside the repo 3. Run `/plugins` 4. Search `Caveman` 5. Install plugin Install once. Works in all sessions after that. ### Manual / Local ```bash git clone https://github.com/JuliusBrussee/caveman.git cd caveman pip install -e . ``` ## Usage — Trigger Commands ### Claude Code ``` /caveman # enable default (full) caveman mode /caveman lite # professional brevity, grammar intact /caveman full # default — drop articles, use fragments /caveman ultra # maximum compression, telegraphic ``` ### Codex ``` $caveman $caveman lite $caveman full $caveman ultra ``` ### Natural language triggers Any of these phrases activate caveman mode: - "talk like caveman" - "caveman mode" - "less tokens please" - "be concise" ### Disable ``` /caveman off # or say: "stop caveman" / "normal mode" ``` Level sticks until changed or session ends. ## Intensity Levels | Level | Trigger | Style | Example | |-------|---------|-------|---------| | **Lite** | `/caveman lite` | Drop filler, keep grammar | "Component re-renders because inline object prop creates new reference each cycle. Wrap in `useMemo`." | | **Full** | `/caveman full` | Drop articles, use fragments | "New object ref each render. Inline prop = new ref = re-render. Wrap in `useMemo`." | | **Ultra** | `/caveman ultra` | Telegraphic, abbreviate everything | "Inline obj prop → new ref → re-render. `useMemo`." | ## Benchmark Results Real token counts from Claude API (reproducible via `benchmarks/` directory): | Task | Normal | Caveman | Saved | |------|-------:|--------:|------:| | React re-render bug | 1180 | 159 | 87% | | Auth middleware fix | 704 | 121 | 83% | | PostgreSQL pool setup | 2347 | 380 | 84% | | Git rebase vs merge | 702 | 292 | 58% | | Async/await refactor | 387 | 301 | 22% | | Docker multi-stage build | 1042 | 290 | 72% | | **Average** | **1214** | **294** | **65%** | > **Important:** Caveman only affects output tokens. Thinking/reasoning tokens are untouched. Caveman make *mouth* smaller, not brain. ## Reproducing Benchmarks ```bash git clone https://github.com/JuliusBrussee/caveman.git cd caveman/benchmarks # Set your Anthropic API key export ANTHROPIC_API_KEY=your_key_here # Run benchmark suite python run_benchmarks.py # Compare normal vs caveman responses python compare.py --task react-rerender python compare.py --task auth-middleware python compare.py --all ``` ## Code Examples — What Caveman Mode Changes ### Before (normal, 69 tokens) ``` The reason your React component is re-rendering is