Ponytail Github

Ponytail Github

Ponytail Github is an open-source AI-coding skill that turns over-eager AI agents into disciplined minimalists. The maintainer Dietrich Gebert modeled the skill on a familiar archetype: the longest-tenured engineer in the office, who looks at fifty lines of code and replaces them with one. Ponytail Github puts that engineer inside your agent of choice.

The skill file is short and human-readable. It instructs the agent to climb a six-rung ladder before writing code: skip the feature if it is not needed, use the standard library if it covers the requirement, use the platform if it ships the feature natively, use an installed dependency if one already does the job, prefer a one-line implementation, and only then write the minimum amount of custom code that actually works. Trust-boundary validation, error handling, security and accessibility are explicitly carved out of the minimization pass and are never removed.

The implementation is equally small. Two Node.js lifecycle hooks (under a hundred lines combined) wire the skill into Claude Code and Codex for always-on activation, while every other supported agent — Cursor, Aider, Continue, Cline, Roo Code, Windsurf, GitHub Copilot, Gemini CLI, OpenCode, Qwen Code, Cody, Amp — receives the skill as a system prompt and stays entirely native. The benchmark suite shipped in the repository runs twelve real feature tickets on the FastAPI full-stack template and reports 54% fewer lines, 22% fewer tokens, 20% lower cost and 27% faster runs at 100% safety on Claude Code 4.5, with reductions up to 94% on over-build traps like date and color pickers. The project is MIT licensed, ships a Spanish translation, and lives at github.com/DietrichGebert/ponytail.