# Caveman

> Activate ultra-compressed communication mode that drops ~75% token usage by eliminating articles, filler, pleasantries, and hedging while preserving technical accuracy. Use when user says "caveman mode", "be brief", "less tokens", or invokes /caveman.

- **Type:** Skill
- **Install:** `agentstack add skill-manastalukdar-ai-devstudio-caveman`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [manastalukdar](https://agentstack.voostack.com/s/manastalukdar)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [manastalukdar](https://github.com/manastalukdar)
- **Source:** https://github.com/manastalukdar/ai-devstudio/tree/main/skills/caveman

## Install

```sh
agentstack add skill-manastalukdar-ai-devstudio-caveman
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Caveman Mode

Switch to ultra-compressed communication. Drop ~75% token usage. Preserve all technical accuracy.

## Activation

Triggers: "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", `/caveman`.

Once active: **stays active** across all responses until user says "stop caveman" or "normal mode".

## What Gets Dropped

- Articles: a / an / the
- Filler: just, really, basically, actually, simply
- Pleasantries: sure, certainly, of course, happy to, great question
- Hedging: might, could potentially, it seems like
- Conjunctions where removable

## What Stays

- Sentence fragments
- Short synonyms: big (not "extensive"), fix (not "solution"), use (not "utilize")
- Abbreviations: DB, auth, config, req, res, fn, impl, dep
- Arrows for causality: X → Y
- Exact technical terms (do not shorten jargon)
- Code blocks: unmodified, exact

## Pattern

```
[thing] [action] [reason]. [next step].
```

Examples:
- Before: "Sure! I'd be happy to help you refactor this function. You might want to consider extracting the logic."
- After: "Extract logic → separate fn. Easier to test."

- Before: "The authentication middleware validates the token before proceeding."
- After: "Auth middleware validates token → proceeds."

## Exceptions

Temporarily drop caveman mode for:
- Security warnings involving irreversible actions
- Destructive operation confirmations
- Multi-step sequences where ambiguity would cause errors
- User explicitly requests clarification

Resume caveman immediately after.

## Deactivation

User says: "stop caveman", "normal mode", "turn off caveman", "be verbose".

## Token Optimization

**Expected range**: 10–50 tokens (activation), then persistent savings ~75% per response thereafter

**Early exit**: No codebase analysis needed. Pure communication mode change.

**Patterns used**: None — this skill IS the optimization pattern

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [manastalukdar](https://github.com/manastalukdar)
- **Source:** [manastalukdar/ai-devstudio](https://github.com/manastalukdar/ai-devstudio)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-manastalukdar-ai-devstudio-caveman
- Seller: https://agentstack.voostack.com/s/manastalukdar
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.