Ponytail

Ponytail is Dietrich Gebert’s cross-agent coding skill/plugin for reducing agent overbuild. Its tagline is ‘the best code is the code you never wrote.’ The p…

Ponytail

Ponytail is Dietrich Gebert’s cross-agent coding skill/plugin for reducing agent overbuild. Its tagline is “the best code is the code you never wrote.” The practical mechanism is a minimality ladder: do not build unnecessary work; prefer stdlib, native platform features, and existing dependencies before writing custom code.

What it is

• A ruleset/skill/plugin for agent harnesses.
• A review lens for unnecessary abstractions and avoidable code.
• A portable adapter set across Claude Code, Codex, OpenCode, Gemini/Antigravity, Pi, Copilot CLI, OpenClaw, Cursor, Windsurf, Cline, Aider, Kiro, Zed, and CodeWhale.

What it is not

• Not a model router.
• Not a replacement for tests or review.
• Not “code golf” if used correctly: safety, validation, security, data-loss handling, and accessibility remain required.

Host integration notes

For OpenCode, the README describes running OpenCode from a Ponytail checkout and adding:

{ "plugin": ["./.opencode/plugins/ponytail.mjs"] }

The plugin injects rules every turn and adds /ponytail commands. OpenCode also auto-loads Ponytail’s AGENTS.md when run from the repo root, so the rules can hold even without the plugin, but mode switching requires the plugin/commands.

For Pi Agent (pi-mono), the install path is:

pi install git:github.com/DietrichGebert/ponytail

For Claude Code and Codex, the source describes plugin marketplace installation plus lifecycle hooks. Node.js must be available in the non-interactive shell path for those hooks; otherwise the skills still work but always-on activation may stay quiet.

Commands

Command	Purpose
/ponytail [lite|full|ultra|off]	Set intensity or report current level
/ponytail-review	Review current diff for over-engineering and return a delete list
/ponytail-audit	Audit the whole repo for over-engineering
/ponytail-debt	Collect deferred ponytail: shortcuts into a ledger
/ponytail-gain	Show measured impact scoreboard from the project benchmark
/ponytail-help	Quick command reference

Adoption stance

Good candidate for our implementation and review agents, especially where AI-Specific Code Pitfalls warns about over-engineering, defensive overreach, and cargo-cult patterns.

Adopt as an optional behavior constraint first, not as a global mandate. Start with implementation/review roles and keep verification gates unchanged.

Related

• Ponytail - source summary and claimed results.
• Model-Task Routing (OpenCode Go) - model selection layer that Ponytail complements.
• Software Design Principles - YAGNI/KISS basis.
• Agent Skills - portable skill mechanism.