Caveman output skill
skillInstruct a compatible agent to answer concisely while preserving code, commands, and error text.
skills/caveman/SKILL.md
Open component record →What JuliusBrussee/caveman is—in plain language
Token-efficiency toolkit for coding agents: a concise-output skill plus a local proxy, recovery store, analyzer, browser, memory, and compression commands.
Mental model: This repository extends a host application. The host loads its commands, skills, hooks, or agents; most day-to-day use happens inside that host rather than by launching a separate app.
InterfacesCLI, local proxy, MCP server, Agent Skill, local report UI
Primary languageGo, TypeScript, Shell, Markdown
Typical useRun caveman learn first, use the output skill independently, and only then evaluate caveman claude/codex wrap mode
Your first 15 minutes
- Complete the linked Installation Guide and its verification step.
- Create a disposable test folder or use non-sensitive sample data.
- Run the small example below and observe what files, ports, or prompts appear.
- Read the result before approving writes, network calls, account access, or costs.
- Only then repeat the workflow with a real project.
Starter workflow
caveman learnComplete indexed functionality map
Each card below corresponds to a component identified in the repository research. Open a component record for its path, purpose, capabilities, dependencies, risks, and relationships.
Agent proxy and wrap
serviceRoute a supported coding agent through a local proxy that applies content-aware request compression.
engine/
Open component record →Learn analyzer
workflowRead local agent history without editing it, rank token sinks, and show a report with suggested fixes.
packages/cli/src/learn-tui.ts
Open component record →Shrink
CLI commandRun a noisy command through recoverable output compression with the caveman shrink command.
packages/cli/
Open component record →Browse
CLI commandExpose a focused, compressed representation of a locally controlled Chrome page.
browse/
Open component record →Durable memory
CLI commandRemember and recall compact facts while retaining a recovery path for original content.
engine/memory/
Open component record →Explore
agentUse a read-only subagent to locate relevant code with compact path-and-line evidence.
packages/agent/
Open component record →TOON encoder
codecRe-encode suitable structured tool results into a smaller representation and decode them when needed.
engine/toon/
Open component record →Pixel mode
workflowRender exceptionally dense text slabs as images when the measured token gate predicts a benefit.
engine/pixel/
Open component record →Caveman MCP tools
MCP serverExpose compression, retrieval, statistics, and TOON operations to an MCP-compatible host.
packages/cli/src/agent-mcp.ts
Open component record →Upstream feature-by-feature guide
These capabilities are derived from the current first-party README and supporting documentation. Names follow upstream terminology so you can search the source documentation precisely.
Upstream capability
See it
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns see it. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
Where your tokens go
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns where your tokens go. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
Caveman Proxy
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns caveman proxy. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
Pixel mode
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns pixel mode. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
Skills as images
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns skills as images. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
The skill
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns the skill. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
Wrap any agent
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns wrap any agent. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
The whole cave
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns the whole cave. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
From inferred to verified
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns from inferred to verified. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
Privacy
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns privacy. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
Upstream capability
Star this repo
Upstream treats this as a distinct part of JuliusBrussee/caveman. In plain language, use this area when your task concerns star this repo. Begin with the documented default, test it on disposable input, and open the source section for its current options and limitations.
How to use the main workflows
Choose the smallest relevant function
Start with one capability card instead of asking the repository to do everything at once. This makes permissions, inputs, and output easier to understand.
Prepare a disposable input
Use a sample URL, copied repository, test document, or sandbox account. Keep production credentials and irreplaceable files out of the first run.
Run, observe, and stop
Watch Terminal output or the host application's activity view. If the behavior differs from the README, stop with Control + C or cancel inside the host before retrying.
Inspect the result
Check generated files, diffs, API responses, logs, or previews. Never assume “command finished” means the result is correct.
Save a repeatable recipe
Record the working command and non-secret configuration in your project's README. Store secrets in the documented environment file or password manager.
Copy-paste recipes from upstream documentation
These examples are selected from the repository's first-party documentation. Replace obvious placeholders, keep quotation marks intact, and run them only in the context indicated by the surrounding explanation.
Open the live upstream manual
open "https://github.com/JuliusBrussee/caveman#readme"Configuration, accounts, and files
Configuration files
Caveman user configuration and local recovery store
A configuration file changes behavior without changing source code. Make one change at a time and keep a backup before editing JSON, TOML, YAML, or environment files.
Important directories
packages/, engine/, skills/
Paths identify where the relevant implementation or generated files live. Paths beginning with ~ are inside your home folder.
Credentials
Existing provider or coding-agent authentication is passed through in wrap mode
Prefer temporary, least-privileged credentials. Never commit .env, tokens, cookies, private keys, or session exports.
External services
Selected LLM provider; Chrome only for browse mode
Check pricing, data retention, rate limits, and account permissions before enabling optional integrations.
Safe operating habits
- Use test data first and keep a current backup or Git commit.
- Read commands before pasting. A README is useful evidence, not a substitute for judgment.
- Review agent-generated file changes with
git diffbefore committing. - Keep local services bound to
127.0.0.1unless you intentionally secure and expose them. - Do not give plugins or MCP servers broader filesystem, browser, GitHub, or cloud access than their current task requires.
- Confirm API costs and model names before running large batches.
- Stop and investigate repeated authentication failures instead of pasting a token into multiple places.
Reference and source trail
The manual reflects first-party files checked on August 18, 2026. It explains the indexed repository rather than promising that every optional third-party integration is available or safe.