How to use the Ponytail Lazy Dev Mode agent skill

What is the Ponytail skill and what does it make the agent do?

Ponytail, described as Lazy senior dev mode, is a skill for any coding task. Its goal is the smallest change that fully solves the problem, written up in a reply a busy person can understand in one read.

Before writing code, the agent is told to read the task and the code it touches, then list every place the change must reach, such as callers, tests, fixtures, config and exports. It then takes the first option that fully works, in this order: skip features nobody asked for, reuse what already exists in the codebase, prefer the standard library or platform features, use an installed dependency, and only then write the minimum code that works.

The skill is deliberately lazy about the solution but not about the change. It still requires you to finish every part the task needs, including the callers, tests and fixtures your change breaks.

• Prefers existing helpers, components and patterns over new code
• Avoids adding dependencies for a few lines of code
• Avoids extra abstractions, options, config and boilerplate
• Adds one small test or assert-based self-check for new non-trivial logic
• Ends each reply with one or two lines on what was skipped or unchecked and what risk to know

How do you install the Ponytail skill?

The skill is distributed as a ZIP file from this directory page. Download the ZIP and unzip it into one of two locations, depending on how you want to use it. The source material for this skill ships as a single file, SKILL.md.

For a personal setup that applies to all your projects, unzip it into ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps rather than a local setup, you can upload the ZIP instead. In the app, go to Settings, then Capabilities, then Skills, and upload it there.

• Personal (all projects): ~/.claude/skills/
• Single project: the project's .claude/skills/ folder
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

How do you trigger Ponytail and switch its level?

Once the skill is installed, it stays active for the whole session. You do not need to invoke it on every request. Ask for a coding task in plain language, such as fixing a bug or adding a small feature to an existing project, and the agent will work in the ponytail style.

The skill has three intensity levels. Full is the default. Lite builds what you asked for and names a smaller option in one line so you can choose. Ultra also questions the request and pushes back on any part the need does not justify before building.

To change the level, use the slash command below. To turn the skill off for the session, say stop ponytail or normal mode.

• lite: build what was asked, and offer a smaller option in one line
• full: the default rules described above
• ultra: question the request first and push back on unjustified parts

What does a typical Ponytail session look like?

Suppose you ask the agent to fix a bug in a function that several parts of your app call. Under this skill, the agent first looks up every caller of that function and then fixes the root cause once in the shared code, rather than patching one call site.

When the work is done, the reply ends with a short note such as what was not tested or what behaviour might change for users. That closing note is the part most teams find useful when reviewing AI-written code.

• Fixes the root cause in shared code instead of adding a workaround at each call
• Keeps the error handling and validation of any code it moves or merges
• Leaves a one-line comment only where the code cannot explain the why
• Labels any shortcut with a known limit so you can see when to upgrade it

Who should use Ponytail, and who should choose something else?

Ponytail suits developers who want small, focused diffs rather than sprawling changes, teams that review AI-written code and want clear notes on what was not checked, and anyone who prefers short, plain-language status replies from a coding agent.

It is a poor fit for tasks that need broad new features or architecture, because the skill deliberately minimises scope and may under-build. The ultra level can even push back on requests. It also assumes an existing project, so it is not useful for non-developers or for anyone without a codebase to work in.

• Good fit: focused bug fixes, small changes in existing projects, reviewable diffs
• Poor fit: large new features, architecture work, non-developers without a codebase

What does Ponytail need, and what are the safety notes?

The skill lists no requirements beyond the agent itself. It does not run scripts and does not use the network, according to its listing. It still needs an existing project to work in.

The skill's guidance says it should not be relied on alone for safety-critical or financial code. Human review and real testing are still needed. The skill also says it must never cut validation at trust boundaries, error handling that prevents data loss, security, accessibility, or anything you asked for.

• Requires: nothing listed beyond an agent and a codebase
• Runs scripts: no
• Uses network: no
• Do not rely on it alone for safety-critical or financial code