lazy-cat plugin
Two Claude Code skills that look for an existing solution before coding and stop additions nobody asked for.
- 50 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skills are instructions. The installer appends a block to `GEMINI.md` and `AGENTS.md`, and the optional hook writes a local token log.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/lazy-catDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skills are instructions. The installer appends a block to `GEMINI.md` and `AGENTS.md`, and the optional hook writes a local token log.
This is third-party code. Review the repository files before installing.
What it does
lazy-cat adds two pause moments to Claude work. The `think-twice` skill fires before choosing an approach and asks whether an API, package or one-liner already exists. The `surgical` skill fires before each code block and asks whether the user requested that code, so extra error handling, tests and abstractions do not appear. The README example is 500 test profiles. Without the skills Claude writes each profile by hand, about 66 thousand tokens, and with them writes a 54-line faker script. The skills install as a plugin, with `npx lazycat-skill` for Claude Code, Gemini CLI and Codex, or as a rule in `CLAUDE.md`. An optional `stats` command counts output tokens per session locally.
Who it is for. Developers who pay for tokens and notice the agent writing from scratch what a library already solves and adding code nobody requested.
Good fit when
- You need 500 test users and the agent plans to write each field by hand
- A phone input or email validation task has a ready library
- A bug fix grows tests and abstractions nobody asked for
- You want rules always in context via `CLAUDE.md` rather than by Claude judgement
- You want to measure token use on your own work with local stats
Not a fit when
- You need a full from-scratch implementation without external dependencies, where `think-twice` points to libraries
- The task needs defensive code, tests and abstractions that you explicitly want
- You expect guaranteed savings: the author shows scenarios with no gain
Example request
Generate 500 staging user profiles and first check whether a ready library exists.Limitations
Savings figures are modeled: output tokens are estimated as characters divided by four across 17 scenarios, and the 500-profile baseline is extrapolated from 15 written profiles. A live check in October 2026 found plain Sonnet already avoided scope creep in small edits, so the gain did not reproduce and cost rose about 8 percent there. In 3 of 17 scenarios `surgical` alone beat both skills combined. Rules for Gemini CLI and Codex are written as a block in `GEMINI.md` and `AGENTS.md`.
How to disable. Disable the `lazy-cat` plugin through `/plugin` or delete the skills from `~/.claude/skills/`. Remove the block between the `lazy-cat:start` and `lazy-cat:end` markers in `GEMINI.md` and `AGENTS.md` manually, and turn stats off with `npx lazycat-skill stats --disable`.
FAQ
How do I install lazy-cat?
Through npm: `npx lazycat-skill` for your whole account or `npx lazycat-skill --project` for the current repository. Through plugins: `/plugin marketplace add albertobarnabo/lazy-cat` and `/plugin install lazy-cat@lazy-cat`.
What do think-twice and surgical do?
`think-twice` fires before choosing an approach and looks for an existing solution. `surgical` fires before each block and removes code nobody asked for.
How do I call a skill manually?
With `/lazy-cat:think-twice <task>` and `/lazy-cat:surgical <task>`.
How do token stats work?
`npx lazycat-skill stats --enable` turns on a hook that logs output tokens per session locally only. `stats` shows totals and trend, and `--disable` removes the hook and keeps the log.
How do I add the rules without skills?
Paste the two rules from the README into the project `CLAUDE.md`. The file is always in context, so applying them does not depend on Claude judgement.
Discussion
No questions or reviews yet.
Related
PluginsNick Tune's plugin marketplace for Claude Code: checks before a task is handed in, code review and TDD personas
A set of 24 slash commands for the Claude Code CLI: commits, review, refactoring, tests and cleanup without repeating the same prompts
Eleven skills for Claude Code and Codex: test audit, git worktree cleanup, plan pages, screen recording transcripts and 1Password sign-in
Claude Code plugins: plan, work and review commands, subagents, hooks and writing skills in a practical minimalism spirit.
Claude Code Blueprint: 12 agents, 19 skills, 16 hooks and 6 CLAUDE.md rules, copy what you need or install a preset
Claude Token Optimizer: fewer tokens at startup
Claude Token Optimizer
The cto utility counts the tokens Claude Code loads at session start and shrinks CLAUDE.md into a short cheat sheet.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| lazy-cat (this card) | no | 50 | |
| Token-Saver | no | 165 | compresses terminal output; lazy-cat changes agent decisions before code is written |
| Token Optimizer | no | 2,964 | cuts tokens across sessions and compactions; lazy-cat is two behaviour rules |
README in short
The README explains the lazy engineer idea, shows a table of modeled token savings across 17 tasks, an honest caveat from a live check and three installation methods. A section covers usage stats.