# JuliusBrussee/cavekit

Frozen — compressed spec-driven development plugin for Claude Code. Still works; active development moved to JuliusBrussee/caveman.

Repository: https://github.com/JuliusBrussee/cavekit
Canonical: https://ross.abutalabs.com/products/cavekit
Homepage: https://caveman.so/
License: MIT
License Family: permissive
Topics: claude-code, parallel-agents, skills, spec-driven-development, test-driven-development
Last push: 2026-08-14T22:07:51+00:00

## Health v2 (maintenance only)
Score: 74/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 97, release rhythm 80, longevity 12
- inputs: {"age_days": 172, "days_push": 19, "days_rel": 137, "gap_med": 6, "n_releases_24m": 4}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1141, forks 86 (observed 2026-08-28T04:03:44.857154+00:00)

## What it is
Cavekit is a frozen (August 2026) Claude Code plugin that implements compressed spec-driven development via a single SPEC.md file and a small command loop (spec, build, check, plus optional grill/research/review/deepen skills). It emphasizes token efficiency through 'caveman encoding' and durable specs that survive context resets, with no sub-agents.

## Use cases
- run spec-driven development in claude code without agent swarms
- keep a durable spec file that survives context resets
- reduce token usage in ai coding workflows
- enforce test-driven development with an ai coding agent
- turn a fuzzy idea into a sharp spec before building
- detect drift between spec and implementation

## When to choose
- you use Claude Code and want a lightweight, token-efficient spec-driven workflow
- you want plan-then-execute discipline without sub-agent overhead
- you want test failures recorded as invariants the spec remembers

## When to avoid
- you need active maintenance, new features, or bug fixes - development moved to caveman
- you don't use Claude Code or a compatible skills-based agent
- you prefer heavyweight multi-agent SDD frameworks with dashboards

## Facets
- artifact type: plugin
- maturity: maintenance
- function: prompt-engineering, agent-framework, testing, documentation, developer-tools
- domain: developer-tools, large-language-models
- platform: cli, cross-platform
- tags: claude-code, spec-driven-development, test-driven-development, token-efficiency, skills, frozen, parallel-agents, automation, ai-agents, nodejs

## Member repositories
- JuliusBrussee/cavekit (main) score 74

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:44.857154+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:35:22.114516+00:00, confidence not recorded.
  - readme: https://github.com/JuliusBrussee/cavekit (fetched 2026-08-28T04:03:44.857154+00:00, sha 1ff850f453b6)
  - homepage: https://caveman.so/ (fetched 2026-08-29T12:40:45.037120+00:00, sha ceb307b7e99b)
  - site_page: https://docs.caveman.so (fetched 2026-08-29T12:40:45.044791+00:00, sha 30c3cb3aee57)
  - site_page: https://docs.caveman.so/docs/engine (fetched 2026-08-29T12:40:45.048917+00:00, sha a8f9999a105b)
  - site_page: https://caveman.so/products/caveman (fetched 2026-08-29T12:40:45.046958+00:00, sha 8bc8aefe4c8e)
  - site_page: https://caveman.so/pricing (fetched 2026-08-29T12:40:45.050847+00:00, sha 1dded1269300)
- Data as of 2026-08-30T08:39:29.467469+00:00.