# Caveman for Claude Code: what it does and how to install it > Caveman is a Claude Code plugin built around an ultra-compressed communication style that cuts output tokens by a measured 65 percent. Beyond that core mode, it adds skills for terse commit messages, one-line code review, task-shaped engineering routines like safe refactors and migrations, three cavecrew subagents for delegated work, and a separate Caveman Cloud layer for tracking LLM spend. Canonical: https://www.flowy.sh/plugins/juliusbrussee-caveman/guide Published: 2026-10-29 Updated: 2026-10-29 Author: Flowy ## What is Caveman? Caveman is a Claude Code plugin built around one idea: talk to the model in a stripped-down register and it spends far fewer output tokens while keeping the same technical content. Flowy indexes the plugin from its public repository, `JuliusBrussee/caveman`, which sits in the development category and carries 100,896 GitHub stars as of the last crawl. Flowy does not host or bundle the code; the real install command lives at [the install card on the listing page](https://www.flowy.sh/listings/juliusbrussee-caveman#install). The plugin ships 30 components: 7 commands, 20 skills, and 3 agents. Six of the seven commands share a name with a skill built for the same job, the command being the slash-triggered form and the skill being what Claude reaches for on its own; `caveman-init` is the one command with no skill twin, since it drops a one-time setup file rather than firing mid-session. ## How does caveman mode compress your replies? The `caveman` skill is the core of the plugin: it cuts output tokens by a measured 65 percent while keeping full technical accuracy, at one of six levels: lite, full by default, ultra, and three further wenyan levels. The `caveman` command sets the level directly or turns the mode off. `caveman-help` is a one-shot reference card for the modes and triggers, and `caveman-stats` reports real token usage for the session by reading the Claude Code session log directly rather than having the model estimate it. `caveman-init` makes the mode persistent: it drops an always-on activation rule into the repository so every IDE agent picks up the same behavior. ## What else does it compress? `caveman-commit` drafts a terse commit message in Conventional Commits format, a subject of 50 characters or fewer, with a body only when the "why" is not already obvious from the diff. `caveman-compress` applies the same idea to a file such as `CLAUDE.md`, a todo list, or a preferences file: it rewrites the file into caveman format while keeping the code, URLs, and structure intact, and keeps a human-readable backup of the original. `caveman-review` shrinks code review the same way, one line per finding covering location, problem, fix, and severity. The `cavecrew-reviewer` agent runs a fuller version of that review over a diff, branch, or file, tagging each line with a severity and skipping nits that don't change behavior. ## What is Cavecrew, and when does it take over? `cavecrew` is a skill, not an agent: it is the decision guide for handing a task to one of the plugin's three agents instead of doing it inline or running a generic explore step. `cavecrew-investigator` is the read-only one, returning a file-and-line table for questions like where something is defined, at around 60 percent fewer tokens than a plain exploration pass, and it refuses to propose fixes. `cavecrew-builder` takes the small, already-located edits: a one-file or two-file change such as a typo fix or a mechanical rename, refusing anything touching three files or more. `cavecrew-reviewer`, covered above, is the third agent and reviews what the builder produced. ## Which skill fits which kind of engineering task? Seven skills route a task by shape rather than compress it. `caveman-explore` is a read-only repository explorer for cold starts and broad cross-file localization. `investigate-first` diagnoses an ambiguous failure or a performance regression before any edit, producing evidence-ranked hypotheses. Once the cause is known, `safe-refactor` restructures code while preserving its behavior, `surgical-patch` fixes a bug at the narrowest layer that can carry it, and `lean-build` handles new feature work under a strict scope and an explicit stop condition. `migration` covers reversible schema, data, API, protocol, configuration, or dependency changes, and `verify-and-stop` is the last step, proving finished work already meets its acceptance conditions without expanding it. ## What is Caveman Cloud? Six more skills wire a project into what the plugin calls Caveman Cloud, a layer for measuring what LLM calls cost. `caveman-setup` connects a repository through the gateway so every request is measured for cost, tokens, and latency with no change in behavior. `caveman-discover` then finds every LLM workflow in that repository and labels it, so spend groups by what the code actually does, such as a support-reply flow or a nightly digest, instead of one anonymous bucket. `caveman-evidence-review` reads the results back, read-only: costs, a Cave Score and Cave Plan, workflows, traces, latency, errors, compression, routing, and verified savings. `caveman-optimize` turns a finding into an operator-chosen candidate change evaluated against a baseline. `caveman-manage` inspects an experiment's eval-gated lifecycle, covering starting, approving, cancelling, promoting, or rolling one back, and reads evidence before acting. `caveman-learn` closes the loop afterward: it reviews the ranked token sinks a report found, applies cost-lowering fixes such as trimming config or moving recurring context to cavemem with consent on each edit, and reports what those fixes returned. Sources: [JuliusBrussee/caveman on GitHub](https://github.com/JuliusBrussee/caveman), [JuliusBrussee/caveman README](https://github.com/JuliusBrussee/caveman#readme) ## Common questions ### What does the caveman skill actually do? The caveman skill switches Claude's replies into a deliberately compressed register that keeps the technical content and drops the filler, cutting output tokens by a measured 65 percent. It runs at one of six levels: lite, full (the default), ultra, and three further wenyan levels that compress harder still. The `caveman` command turns it on or off and lets you pick the level directly. ### Does caveman mode lose information, or just make replies shorter? The plugin's own description says the compression targets filler, not content: it is built to cut output tokens by a measured 65 percent while keeping full technical accuracy. It also does not ask you to take that number on faith. The `caveman-stats` skill and command read token counts straight from the Claude Code session log, and the description is explicit that the model itself does not compute the savings. ### What is Cavecrew, and how is it different from caveman mode? Cavecrew is a separate delegation layer built from three agents: `cavecrew-investigator` for read-only code location, `cavecrew-builder` for a small, already-located edit, and `cavecrew-reviewer` for diff and PR review. The `cavecrew` skill is the decision guide that tells the main thread when to hand work to one of those three agents instead of doing it inline. It saves tokens on exploring and editing a codebase, which is a different problem from caveman mode compressing ordinary replies. ### Does Caveman only work inside a single session, or does it track LLM cost too? Both. Most of the plugin's skills and commands work inside one Claude Code session: compressing replies, drafting commits, reviewing diffs, and choosing which engineering skill fits a task. A separate set of skills wires a project into what the plugin calls Caveman Cloud: `caveman-setup` connects the gateway that measures every LLM request, `caveman-discover` labels the workflows that spend runs through, and `caveman-evidence-review`, `caveman-optimize`, `caveman-manage`, and `caveman-learn` cover reading that evidence, proposing a fix, running its experiment lifecycle, and applying an approved fix. ### Which skill should I use when I don't know what's broken yet? That is what `investigate-first` and `caveman-explore` are for. `investigate-first` diagnoses an ambiguous failure before you start editing, ranking hypotheses by evidence rather than guessing. `caveman-explore` is the read-only repository explorer for cold starts, when you need to find where something lives rather than why it is failing. Once you know what has to change, `safe-refactor`, `surgical-patch`, `lean-build`, and `migration` are the skills that shape how the edit itself gets made. ## Related - [The listing page](https://www.flowy.sh/listings/juliusbrussee-caveman): Install command, components and repository - [Alternatives to this plugin](https://www.flowy.sh/alternatives/juliusbrussee-caveman): Peers in its category, ranked by stars - [The Claude Code plugin marketplace](https://www.flowy.sh/marketplace/plugins): How plugins install, and every plugin in the index - [superpowers for Claude Code: what it does and how to install it](https://www.flowy.sh/plugins/superpowers-flow/guide): another plugin guide - [mattpocock/skills for Claude Code: install and use](https://www.flowy.sh/plugins/matt-pocock-flow/guide): another plugin guide