Ponytail 🧩 Skill Open source

What it does

Ponytail is a skill and plugin that makes a coding agent write only the code a task actually needs. Before writing anything, the agent works down a fixed ladder of questions. Does this need to exist at all? Is it already in the codebase? Does the standard library or the platform already do it? Is there an installed dependency that handles it? Does it fit in one line? Only after those checks does it write the smallest version that works. A typical result is a native <input type="date"> instead of a new date-picker library. The rules say validation, error handling, security and accessibility must never be cut. The author's benchmark on a FastAPI and React repository reports much smaller diffs and lower cost than the same agent without the skill.

What is inside

• Always-on ruleset with lite, full, ultra and off levels, set by command, environment variable or config file
• /ponytail-review: flags over-engineering in the current diff and returns a list of things to delete
• /ponytail-audit: runs the same check on the whole repository
• /ponytail-debt: collects deferred ponytail: shortcuts into a ledger
• /ponytail-gain and /ponytail-help
• Small Node.js lifecycle hooks that switch the rules on automatically and pass them to subagents

Works with

Plugins or extensions exist for Claude Code, Codex, GitHub Copilot CLI, Gemini CLI, Antigravity CLI, OpenCode, Devin CLI, Cursor (through hooks) and several others. Windsurf, Cline, Aider, Kiro, Zed, JetBrains Junie, Amp and Jules can use the included AGENTS.md or rule files, but only as instructions.

How to install

In Claude Code, send these as two separate prompts:

/plugin marketplace add DietrichGebert/ponytail
/plugin install ponytail@ponytail

Maintenance and safety

The project is MIT licensed and actively maintained, with a test suite and benchmarks you can reproduce. The hooks run node scripts on each session, and Codex asks you to trust them first. Uninstalling leaves a few small state files, which the README lists. Because it pushes for less code, review what it leaves out.

Who should use it

Developers whose agents keep adding dependencies, wrappers and abstractions, and who want smaller diffs that are easier to review.