dsh-ponytail

When using AI-assisted coding, we often face a dilemma: models tend to generate feature-complete but potentially over-engineered solutions, while what we need is concise, just-enough implementations. The dsh-ponytail plugin is designed to solve this exact pain point. It deeply adapts the upstream Ponytail minimalist coding philosophy into DeepSeek Harness, injecting a structured "lazy developer" rule set into every model interaction to fundamentally guide the AI toward producing simpler, YAGNI-compliant code.

The plugin's core capability lies in its session-level "YAGNI Decision Ladder" and flexible mode system. Users can switch between the lite, full, or ultra modes with simple commands based on the task at hand. These modes are not mere labels; they represent different intensities of minimalist decision logic, ranging from "deliver exactly what's explicitly requested" to "actively question speculative features," systematically challenging unnecessary complexity while ensuring safety and correctness. Furthermore, it offers one-shot skills (like /ponytail-review) for reviewing recent changes, auditing entire repositories for over-engineering, or cataloging technical debt, forming a closed loop of pre-guidance and post-review.

This plugin is ideal for development teams and individuals who prioritize code quality and efficiency, especially in scenarios involving frequent AI collaboration and a desire to control technical complexity. If you're tired of repeatedly emphasizing "please keep it simple" in prompts or struggle to clean up redundant AI-generated code, dsh-ponytail offers a systematic solution. It transforms "keeping it simple" from a slogan into a configurable, executable, and workflow-integrated habit throughout your development session.