Ponytail

// Overview

Ponytail is an agent skill and plugin whose whole job is to stop a coding agent from over-building. Its persona is the senior developer who looks at your fifty lines, says nothing, and replaces them with one. In practice it installs a ruleset — plus a handful of slash commands — into whichever coding agent you already use, and that ruleset runs before the agent writes code.

The rule is a seven-rung ladder, and the agent stops at the first rung that holds: does this need to exist at all (YAGNI), is it already in the codebase, does the standard library do it, is there a native platform feature, is it in an installed dependency, can it be one line, and only then the minimum that works. The README is explicit that the ladder runs after the agent understands the problem, not instead of it — lazy about the solution, never about reading the code.

It is also explicit about what is never cut. Trust-boundary validation, data-loss handling, security and accessibility stay in, so "less code" means less unnecessary code rather than golfed code. The repository publishes its own benchmark to back the claim: twelve feature tickets against a real FastAPI + React template with a headless Claude Code session, scored on the resulting git diff. Against the same agent with no skill, it reports -54% lines of code, -22% tokens, -20% cost and -27% time at 100% on its adversarial safety tier — and it documents that an earlier single-shot benchmark overstated the effect, with the agentic numbers as the corrected version.

Installation is per-agent rather than per-project: the same repository ships a Claude Code and Codex plugin marketplace, a Copilot CLI plugin, a Gemini CLI / Antigravity extension, an OpenCode plugin entry, Hermes, Devin, Swival, Pi and OpenClaw paths, and an AGENTS.md fallback that agents such as CodeWhale and Qoder pick up with no setup at all. Levels (lite, full, ultra, off) switch how aggressively the ruleset is applied.

// What it does

• A seven-rung decision ladder applied before code is written: YAGNI, reuse, stdlib, native platform feature, installed dependency, one line, then minimum that works
• Safety carve-outs that are never traded away — validation at trust boundaries, data-loss handling, security and accessibility
• Levels (lite / full / ultra / off) plus commands such as /ponytail-review, /ponytail-audit, /ponytail-debt and /ponytail-gain
• Installs into 20+ agents: Claude Code, Codex, Copilot CLI, Gemini CLI / Antigravity, OpenCode, Hermes, Devin, Swival, Pi, OpenClaw and more
• An AGENTS.md fallback so agents that auto-load repo instructions get the ruleset with zero configuration
• A published, reproducible benchmark suite (agentic and single-shot) with per-task tables and stated limitations

// Getting started

Ponytail installs as a plugin or skill inside your agent, not as a dependency of your project. The Claude Code and Codex plugins run two small Node.js lifecycle hooks, so node must be on your PATH for always-on activation; without it the skills still work.

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

codex plugin marketplace add DietrichGebert/ponytail
codex plugin add ponytail@ponytail

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

gemini extensions install https://github.com/DietrichGebert/ponytail

# opencode.json
# { "plugin": ["@dietrichgebert/ponytail"] }

hermes plugins install DietrichGebert/ponytail --enable

/ponytail lite
/ponytail ultra
/ponytail-review

npx promptfoo eval -c benchmarks/promptfooconfig.yaml

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Reach for it when your agent answers "add a date picker" with a new dependency, a wrapper component and a stylesheet
• Use it on an existing codebase via /ponytail-review or /ponytail-audit to find code that should not have been written
• Add it to a team's agent setup through AGENTS.md so every harness picks up the same restraint rules
• Pick it when agent token spend and latency matter — the project's own agentic benchmark reports lower cost and time alongside smaller diffs

// How Ponytail compares

Ponytail alongside other open-source agent skills & plugins tools AI/TLDR tracks, ranked by GitHub stars.