Caveman Skills - Setup Guide¶
Source: juliusbrussee/caveman (2.2M combined installs) GitHub: github.com/juliusbrussee/caveman Category: Code Quality / Agent Workflow First Seen: August 12, 2026 Quality Tier: 🟢 Production (seven skills above 250K installs)
Caveman is a blunt-instrument agent coding workflow: keep it simple, commit everything, compress the context, review with evidence. The flagship caveman skill alone has 431K installs, backed by a full toolchain - caveman-commit, caveman-review, caveman-compress, caveman-stats, and the multi-agent cavecrew.
Installation¶
# Install the full repo
npx skills add juliusbrussee/caveman
# Or install the core individually
npx skills add juliusbrussee/caveman --skill caveman
npx skills add juliusbrussee/caveman --skill caveman-commit
npx skills add juliusbrussee/caveman --skill caveman-review
npx skills add juliusbrussee/caveman --skill caveman-compress
Core Skills¶
| Skill | Installs | Use For |
|---|---|---|
caveman |
431.1K | Main workflow - simple, direct coding loops |
caveman-commit |
303.1K | Disciplined commit cadence |
caveman-review |
301.0K | Evidence-based code review |
caveman-compress |
298.8K | Context compression between steps |
caveman-help |
293.1K | Self-help and usage discovery |
cavecrew |
250.6K | Multi-agent crew orchestration |
caveman-stats |
250.3K | Progress and usage statistics |
compress |
46.6K | General context compression utility |
Smaller but useful additions: caveman-explore, caveman-optimize, investigate-first, verify-and-stop, caveman-learn, surgical-patch, safe-refactor, caveman-discover, lean-build, caveman-manage, caveman-evidence-review, caveman-setup, migration, plus caveman-es (Spanish) and caveman-cn (Chinese) localizations.
Prerequisites¶
| Requirement | Details |
|---|---|
| Node.js + npx | For the skills.sh CLI install path |
| Git repo | Workflow is commit-centric |
| Any agent runtime | Skills are runtime-agnostic; Hermes/Claude Code-style agents work |
CorpusIQ Use Cases¶
| Use Case | How |
|---|---|
| Docs repo maintenance | caveman-review + caveman-commit for sweep-driven doc updates like this catalog |
| Context discipline | caveman-compress pattern matches the token-optimization doctrine for long agent sessions |
| Multi-agent builds | cavecrew when parallel workers need a shared coding protocol |
| Safe refactors | surgical-patch + safe-refactor for production repo edits |
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.
Limitations / Verification¶
- Opinionated workflow - best adopted whole, not piecemeal
- Verify install:
npx skills list | grep cavemanshows installed entries caveman-es/caveman-cnare localizations of the core skill
Related¶
Powered by CorpusIQ