lean-build Skill for Claude: Build Features Without Overbuilding

lean-build makes your coding assistant build a feature as a complete narrow path, reusing your code and stopping at the acceptance line.

Installs: 111.3K; on skills.sh
Stars: 116.7K; on GitHub
Files: 1; besides SKILL.md
Updated: Nov 2026; last repo commit

lean-build tells the assistant to turn a feature request into acceptance conditions and explicit non-goals taken from the request and the repository. It traces the entry point through the layers that own the rules and delivers one coherent end-to-end path. It reuses a fitting existing seam, and leaves out modes, providers, config, extensibility and polish unless the acceptance needs them.

It loads for new behavior, product slices and integrations where overbuilding is a risk. It exercises the path, runs focused proof, stops when acceptance passes, and reports only material omissions. The SKILL.md opens with the line "Native Core's architecture-first simplicity remains mandatory" and never says what Native Core is. It bundles no scripts.

Good for

Developers adding a feature to an existing codebase
Teams that want small diffs and no speculative options
Reviewers tired of oversized AI-written changes

Install lean-build

Claude Code

Copy this prompt.

Install the lean-build skill in Claude Code. Run this command:

npx skills add https://github.com/juliusbrussee/caveman --skill lean-build

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

Build feature work with high overbuilding risk. Use for new behavior, product slices, or integrations where repository reuse, strict scope, and an explicit stop condition matter.

Key facts

Author: Julius Brussee (repo owner juliusbrussee)
Repository: juliusbrussee/caveman, folder skills/lean-build
Installs on skills.sh: 111,277 (as of 2026-11-02)
Repo stars: 116,702 (whole repository, as of 2026-11-02)
License: Apache-2.0
Last repo commit: 2026-11-01
Files in the skill folder: 2 files: SKILL.md and agents/openai.yaml
skills.sh audits: Gen Agent Trust Hub: pass, Socket: pass, Snyk: pass

What it includes

Write down what counts as done and what is out
Derives observable acceptance and explicit non-goals from the request and the repository.
Read · Acceptance and non-goals
Follow the entry point through layers that own the rules
Traces the entry point through the layers that own the invariants.
Read · Trace the layers
Deliver one end-to-end path, not a one-file hack
Delivers a coherent end-to-end path across the responsible layers rather than forcing the work into one file.
Write · One complete path
Reuse what fits; refactor when patching would duplicate
Reuses a fitting seam and refactors when patching would duplicate behavior or weaken ownership.
Reuse the seam
No modes, providers, config or polish unless required
Omits modes, providers, config, extensibility and polish unless the acceptance needs them.
Leave things out
New dependency, service or migration only with a stated reason
Adds surface, a dependency, a service, config or a migration only for lifecycle design or acceptance, and states the material tradeoff.
Justify new surface
Run the path, run focused proof, stop at acceptance
Exercises the path, runs focused proof, stops when acceptance passes, and reports only material omissions and the trigger.
Runs code · Exercise and stop

From the official docs

Permissions

Low risk

The skill has the assistant write code and run checks, so the risk is a wrong or oversized edit, not a data leak.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It adds code and may add dependencies
It is told to avoid new dependencies, services and config unless acceptance needs them, but it can add them. Review the diff and lockfile changes.

Questions

What does the lean-build skill do?

It makes the assistant build a feature as one narrow, complete path. It sets acceptance and non-goals first, reuses what the repo already has, and stops when acceptance passes.

How do I install lean-build in Claude Code, the Claude app or Codex?

Use the install box at the top of the page and pick your client. The SKILL.md bundles no scripts and lists no other requirements. The folder also includes an agents/openai.yaml file. The README says your agent picks these work patterns up when a task fits, so you do not need to call it by name.

Does lean-build need an API key or an account?

No. The SKILL.md asks for no key, login or tool. It is plain instructions for the assistant, and it bundles no scripts.

Is lean-build safe to use?

It is a short instruction file and bundles no scripts. What it changes is how the assistant behaves when it edits your project, so review the diff and keep your usual git and test setup. The permissions section below lists what to watch.

How is lean-build different from surgical-patch?

lean-build is for new behavior, product slices and integrations. surgical-patch is for bug fixes at the layer that owns the bug. Use lean-build when you are adding something and want the assistant to keep it small.

What is the "Native Core" the SKILL.md mentions?

The SKILL.md says "Native Core's architecture-first simplicity remains mandatory" and later says to "preserve Core safety", but it does not define either term. The README does not explain them either, so we cannot say what they refer to.

Information

updated Nov 2, 2026

Repository: juliusbrussee/caveman
Folder: skills/lean-build

juliusbrussee/caveman skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 4

lean-build SKILL.md · github.com
caveman README · github.com
lean-build openai.yaml · github.com
surgical-patch SKILL.md · github.com