Agent Skills

spec

A Claude Code plugin that turns natural language into blueprints, blueprints into parallel build plans, and build plans into working software with automated iteration, validation, and cross-model peer review.

Instalação
CLI
npx skills add https://github.com/juliusbrussee/cavekit --skill spec

Instale esta skill com a CLI e comece a usar o fluxo de trabalho SKILL.md em seu espaço de trabalho.

Última atualização em 4/23/2026

cavekit

compressed spec-driven development for claude code
one file · three commands · zero sub-agents

what this is

Plan-then-execute forgets. SDD remembers — but most SDD frameworks bury
that value under agent swarms, dashboards, and ceremony that costs more
tokens than it saves.

Cavekit 4 is a rewrite from the ground up. It keeps only what earns its
place:

  • durable spec — SPEC.md at repo root survives context resets.
  • caveman encoding — ~75% fewer tokens than prose. Symbols, fragments,
    pipe tables for repeating records.
  • backprop reflex — every test failure becomes a §B entry; classes
    of bug become §V invariants the spec never forgets.

That's the whole pitch.

commands

cmd job
/ck:spec create / amend / backprop SPEC.md. Sole mutator.
/ck:build native plan → execute against spec. Auto-backprops on failure.
/ck:check read-only drift report. Lists §V / §I / §T violations.

install

One line, via the skills CLI:

npx skills add JuliusBrussee/cavekit

Installs five skills into ~/.claude/skills/: spec, build, check
(the workflow) plus caveman and backprop (the utilities). Claude
activates each when its trigger context matches — e.g. "write a spec
for…" invokes spec, "build the next task" invokes build. Claude Code
picks them up on next launch.

Or via the Claude Code marketplace (also adds /ck:spec, /ck:build,
/ck:check slash commands):

/plugin marketplace add juliusbrussee/cavekit
/plugin install ck@cavekit

Or clone directly:

git clone https://github.com/juliusbrussee/cavekit.git ~/.claude/plugins/cavekit

format

See FORMAT.md. Fixed sections: §G goal, §C constraints,
§I interfaces, §V invariants, §T tasks (pipe table), §B bugs (pipe table).

files

FORMAT.md             spec schema + caveman encoding rules
commands/             three slash-command entry points (/ck:spec, /ck:build, /ck:check)
skills/spec           spec mutator (mirrors commands/spec.md as a skill)
skills/build          plan-execute skill (mirrors commands/build.md)
skills/check          drift report skill (mirrors commands/check.md)
skills/caveman        encoding utility
skills/backprop       bug → spec protocol (six steps)

non-goals

  • no sub-agents. Main Claude does the work.
  • no dashboards. cat SPEC.md is the dashboard.
  • no parallel workers. One thread, one spec, one diff.
  • no JSON / YAML spec bodies. Markdown + pipe tables.
  • no hooks, no orchestration binaries, no TypeScript helpers.

older cavekit (the Hunt lifecycle, v3.1.0 and earlier)

The previous generation is not deprecated — it is frozen at tag
v3.1.0 and
remains a fully working plugin.

What it is:

Spec-driven AI development with an autonomous execution loop. Four-command
Hunt lifecycle (/ck:sketch → /ck:map → /ck:make → /ck:check),
plus /ck:ship, /ck:review, /ck:revise, /ck:status, /ck:design,
/ck:research, /ck:init, /ck:config, /ck:resume, /ck:help — 16
slash commands total. 12 named sub-agents. Per-task token budgets,
stop-hook state machine, model-tier routing, auto-backpropagation from
test failures, tool-result caching, Codex peer review, Karpathy
behavioral guardrails, caveman token compression, knowledge-graph
integration, and design-system enforcement. Parallel wave execution and
team mode.

Pick v3.1.0 if you want the full autonomous loop, parallel agents,
peer review, or design-system workflow. Pick v4 if you want the
distilled core — one spec, three commands, no orchestration.

install the older version

Marketplace:

/plugin marketplace add juliusbrussee/[email protected]
/plugin install ck@cavekit

Git:

git clone -b v3.1.0 https://github.com/juliusbrussee/cavekit.git

Full docs live at the tag — git checkout v3.1.0 and read the README
there for command reference, skill catalog, and the Hunt lifecycle guide.

choosing, or moving

See UPGRADE.md. Honest framing:

  • Stay on v3.1.0 if your project has active context/kits/ investment.
  • Move to v4 if you want fewer moving parts and smaller token bills.
  • It is a two-way door — SPEC.md is plain markdown; nothing traps
    you in either direction.

philosophy

The spec is the only artifact that earns its tokens. Everything else
that costs tokens must either save more tokens later, or the user's
attention, or it gets cut.

See CHANGELOG.md for the full v3 → v4 break.

license

MIT.