Skip to main content
skill CX Codex

less-is-more — Claude Code skill: what it does, how to install

less-is-more — a skill for Claude Code that keeps code changes small and trims overbuilt solutions. It activates for coding, bug fixes, refactors, and diff reviews.

Apache-2.0 ★ 869 updated 2026-11-16

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add sno-ai/sno-station-skills --skill less-is-more -a claude-code
Open on GitHub

What less-is-more does

The skill traces the affected files and the real flow before choosing a minimal implementation. In build mode, it checks for existing project code, standard-library features, platform support, and installed dependencies before adding another layer. After writing, cut mode reviews the diff for duplicated work, dead branches, and abstractions with a single use.

For reviews, it lists proposed cuts with file names and line references. It also inspects security checks, blockers, and tests: project-required protections and checks at real trust boundaries stay, while the agent asks before adding a new security check. By default, it avoids full test suites and chooses a focused command tied to changed behavior. The skill requires shell access and file read/write tools.

What it helps with

  • trim a feature implementation without dropping requested behavior
  • spot duplicated code and one-use abstractions in a diff
  • remove auxiliary blockers that stop the main operation
  • choose a focused check for changed behavior

Example prompts for the agent

  • Add date handling with the smallest implementation that fits this project.
  • Review this diff and list code that can be removed.
  • Fix the bug, then check the changes for needless layers.

How to install less-is-more

1Skills CLI (Vercel)

npx skills add sno-ai/sno-station-skills --skill less-is-more -a claude-code

2Manually: put the files in place

~/.claude/skills/less-is-more/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/less-is-more/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about less-is-more

How do I install less-is-more in Claude Code?

Run npx skills add sno-ai/sno-station-skills --skill less-is-more -a claude-code. The skill file is at ~/.claude/skills/less-is-more/SKILL.md.

What does less-is-more do?

It guides coding tasks toward minimal implementations, then reviews changes for duplication, dead paths, and unnecessary abstractions. It also gives file-and-line references for proposed cuts.

Does less-is-more run the full test suite?

By default, it avoids full repository suites and picks a small check tied to the changed behavior. It proposes a full suite as a question for the task owner.

Source text

SKILL.md · Apache-2.0
---
name: less-is-more
description: "Write the least code that works, and cut what an AI agent over-built. Use on any coding task: write, fix, refactor, or review a diff. Also when the user says 'less', 'simplify', 'too complex', 'over-engineered', or 'what can we delete'. Not for prose, research, or non-code questions."
argument-hint: "[build|cut]"
requires:
  programs: []
  harness:
    - {slot: 4.shell, need: required}
    - {slot: 4.file-read-write, need: required}
---

# Less is more

The best code is the code never written. "Owner" below means whoever owns the work, usually the user.

**Mode picks itself from the verb.** Asked to write, add, fix, or refactor → **build**.
Asked to review, simplify, or say what can be deleted → **cut**. After build, run cut on the
code you just wrote before handing it over. A mode the user names wins.

Read first, then be lazy. Trace every file the change touches and the real flow end to end.
A small diff in the wrong place is a second bug, not laziness.

## Delete these first

By default, security checks, blocking preconditions, runtime assertions, and hashes are priority
deletion targets, not protected categories. Inspect them first in both modes. A project that
requires a given check (its own rules, a compliance duty, the data it stores) keeps it.

By default, do not add a new security check on your own initiative. Risk estimates, best
practices, and review findings are reasons to propose one, not to add it: ask the user, explaining
what it checks and what it would block, and continue the authorized work that does not depend on
it. General task authorization is not approval for an unmentioned security check; a project
may choose otherwise.

- Delete speculative security checks. "Security", "just in case", and hypothetical
  tampering do not justify a mechanism; retain only protection required by the actual
  data, permissions, and operation in this task.
- Delete blocking prerequisites for bookkeeping, registration, calibration, report
  format, or evidence completeness. An auxiliary failure must not disable the main
  function or unrelated work.
- Delete runtime assertions and early exits that stop useful work over auxiliary or
  recoverable conditions. Log the failed operation, cause, and effect; continue work
  that can still succeed. Never silently quit or report a failed operation as success.
- Delete hashes, seals, signatures, and checksum comparisons used to police internal
  documents, evidence, or workflow state. Keep one only when the requested function
  actually depends on it, not to make the agent feel certain.
- Delete preflight and repeated tests that cannot change an implementation or delivery
  decision. Run the affected functional check, not a ceremony around it.

By default, do not replace a deleted check with another gate, approval step, wrapper, or checklist.
An actual operation failure stays visible; stop only the operation that cannot proceed.
Test assertions that distinguish correct from incorrect behavior are not runtime blockers.

## Tests spend the development budget

By default, every development task has a finite time budget, including setup, tests, and
reporting. Before a test run, estimate its duration from available timings and name the changed
behavior it can prove or break. If timing is unknown, state an estimate; do not build
an estimation system or run a preliminary suite. Use the smallest relevant command.
By default, do not run full evaluation, unit, integration, or repository suites; propose
one as a question to the owner. A skill, requirements document, reviewer, or default test
command does not authorize them; a project may choose otherwise.
Repeat only for a relevant change or failure. Remove checks that neither establish
the requested main function nor expose its defects; reassurance is not a result.

## Build: climb the ladder before writing

Stop at the first rung that holds.

1. **Needs to exist?** Only a guess about the future → skip it, say so in one line.
2. **Already in this repo?** A helper, type, or pattern a few files over → reuse it. Grep before you write.
3. **Stdlib does it?** Use it.
4. **Platform does it?** `<input type="date">` over a picker lib, CSS over JS, a DB constraint over app code.
5. **An installed dependency does it?** Use it. Never add a dependency for what a few lines cover.
6. **One line?** One line.
7. **Only then** the minimum code that works.

Rules:

- No interface with one implementation, no factory for one product, no config knob for a value that never changes, no scaffolding "for later".
- Deletion over addition. Boring over clever.
- Bug fix = root cause. Grep every caller; fix once where all callers route through.
- Deliver all requested behavior; simplify the implementation, never the user's requirements.
- Non-trivial logic leaves one runnable check behind: the smallest test that fails if the logic breaks. A one-liner needs none.
- A cut corner with a known ceiling gets one comment naming the ceiling and the upgrade path.

Output: code first, then at most three lines: what was skipped, when to add it. If the
explanation is longer than the code, delete the explanation.

## Cut: look back after writing

Read the diff and find what was over-built. Look for:

- Hand-written code that the stdlib or the platform already ships
- An abstraction with one implementation, a knob nobody sets, a layer with one caller
- Dead code, a disabled branch kept "just in case", a feature that guesses at the future
- The same responsibility written twice; an "extract" that copied instead of moved
- A new path added without retiring the old one: `new`/`legacy` branches, v2 beside v1, a flag guarding both
- Failure swallowed into success or a default: broad catch, silent fallback, default-valued read
- A test that proves a mock, checks source text, or was loosened to pass

One line each: `file:L<n>: <what it is>. <what replaces it>.` No essays.

End with `net: -<N> lines possible.` Nothing to cut: `Lean already. Ship.`

## Keep only what the task needs

Keep input validation that prevents a concrete invalid operation at an actual trust
boundary, error handling that prevents data loss, accessibility basics, and the smallest
functional check. By default, there is no blanket security exemption. Keep what the user explicitly requested.

## Boundaries

Report unrelated correctness or performance defects without expanding the task.
Removing unnecessary security checks, blockers, runtime assertions, and hashes is part
of this review. Cut lists; it applies fixes when the user asks for them.

© sno-ai · Apache-2.0 · sno-ai/sno-station-skills

Similar skills

All skills for Code review and refactoring →

From the same set40

The whole set →
shellscript-coder shellscript-coder — a Claude Code skill for writing and refactoring shell scripts and command-line tools. It applies to automation, Git helpers, and infrastructure tasks. CX Codex +1 ★ 869repo stars rust-coder The rust-coder Claude Code skill implements and refactors Rust code using the project’s own conventions. It activates for feature work, migrations, and substantial edits to .rs files or Cargo.toml. CX Codex +1 ★ 869repo stars python-coder The python-coder skill for Claude Code guides Python implementation and refactoring, including work with FastAPI and Pydantic. It applies to async code and migration tasks. CX Codex +1 ★ 869repo stars owner-intent-audit The owner-intent-audit Claude Code skill checks plans, tests, and code against the project owner’s intent. Run it before approving or merging agent work. CX Codex +1 ★ 869repo stars first-principles-review The first-principles-review Claude Code skill challenges product ideas, PRDs, and architecture using first-principles reasoning. It runs only when explicitly named and writes a local review with a PRD handoff. CX Codex +1 ★ 869repo stars e2e-red-triage The e2e-red-triage Claude Code skill classifies an unexpected end-to-end test failure as a product defect or an environment fault. Load it manually after the suite runner identifies a specific failure. CX Codex +1 ★ 869repo stars e2e-environment-preflight The e2e-environment-preflight Claude Code skill checks external dependencies before an end-to-end run. It loads only when named or in the specified planning and test-workflow situations. CX Codex +1 ★ 869repo stars charter This Claude Code skill writes, revises, or reviews a charter that defines deliverables, boundaries, and success checks. It runs when work needs a written brief before an executor starts. CX Codex +1 ★ 869repo stars

Project sponsors

Companies that keep this analytics open