Ponytail is a Node-based, locally runnable skill for AI coding assistants that applies a reuse-first decision ladder before new code is written. On skill-capable hosts, it also supports diff review and repository audits that produce a delete-list of unnecessary code; other adapters load only the always-on ruleset.
Project overview
Ponytail is a Node-based, locally runnable skill for AI coding assistants that applies a reuse-first decision ladder before new code is written. On skill-capable hosts, it also supports diff review and repository audits that produce a delete-list of unnecessary code; other adapters load only the always-on ruleset.
Project type
AI Agent · AI Coding · Prompt Engineering
Use cases
Coding & Development
Deployment
Refer to project documentation
License
MIT
Best for
Developers using a supported AI coding assistant who want agents to favor existing code, standard libraries, and native platform features before adding implementation.
Teams that want to inspect a current Git diff or repository for over-engineering while retaining validation, security, accessibility, and data-loss safeguards.
Key capabilities
Instructs agents to evaluate whether work is necessary, then consider existing code, standard libraries, and native platform features before writing new code.
Retains existing safeguards for trust boundaries, data loss, security, and accessibility during code reduction.
Provides commands to review the current Git diff or audit an entire repository for over-engineering and produce a delete-list.
Supports lite, full, ultra, and off rule intensity through commands or environment variables.
Limitations and risks
Commands and slash functions work only on skill-capable hosts. Other adapters load the always-on ruleset without those commands.
For lifecycle hooks to work, Node.js must be available on the non-interactive shell PATH.
Getting started
For the documented Claude Code path, send `/plugin marketplace add DietrichGebert/ponytail`, then send `/plugin install ponytail@ponytail`.
Evidence and sources
GitHub project description: Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
README: You ask for a date picker. Your agent installs flatpickr, writes a wrapper component, adds a stylesheet, and starts a discussion about timezones.
README: Works with 20 agents
README: ~54% less code (up to 94%) · ~20% cheaper · ~27% faster · 100% safe
README: Before writing code, the agent stops at the first rung that holds: ``` 1. Does this need to exist? → no: skip it (YAGNI) 2. Already in this codebase? → reuse it, don't rewrite 3.…