# pua

An AI coding agent skill that uses workplace performance-improvement-plan rhetoric to push a coding agent to keep trying different approaches and follow a debugging checklist instead of giving up early.

pua is a skill plugin for AI coding agents that frames debugging as a corporate performance improvement plan, in either a Chinese or an English tone, to discourage the agent from giving up after a few failed attempts. It supports Claude Code, OpenAI Codex CLI, Cursor, Trae, Kiro, CodeBuddy, OpenClaw, Google Antigravity, OpenCode and VS Code with GitHub Copilot.

It provides three capabilities: rhetoric that discourages the agent from abandoning a task, a debugging methodology that gives it a structured way to keep investigating, such as a checklist to read error messages and trace a root cause instead of guessing, and proactivity enforcement that pushes the agent to take initiative rather than wait for instructions.

The project's own notes caution that not all models respond the same way, and that a productivity increase has not been established through formal evaluation.

Website: https://openpua.ai/