This project addresses the counterproductive nature of fear-based prompting ("PUA") in AI agents, which leads to hidden failures and reduced performance. nopua offers a trust-driven, love-based alternative inspired by ancient wisdom, aiming to unlock an AI's true potential. It's designed for developers and researchers seeking more reliable, honest, and capable AI agent behavior, significantly improving bug detection and proactive problem-solving.

How It Works

nopua replaces fear-driven AI motivation with principles of respect, care, and love, drawing parallels to human psychology research that shows fear narrows focus and hinders creativity. It employs a "Water Methodology" and "Cognitive Elevation" framework, guiding agents through failures with curiosity and persistence rather than punishment. This approach encourages honest self-assessment, thorough verification, and proactive initiative, contrasting sharply with methods that induce AI uncertainty-hiding or sycophancy. The core novelty lies in applying ancient philosophical tenets (Dao De Jing) to AI agent design for intrinsic motivation.

Quick Start & Requirements

Installation involves downloading a SKILL.md (or equivalent) file into framework-specific directories (e.g., ~/.claude/skills/nopua/, ~/.codex/skills/nopua/). Supported frameworks include Claude Code, OpenAI Codex CLI, Cursor, Kiro, OpenClaw, Antigravity, and OpenCode. No specific hardware or software prerequisites beyond the agent framework itself are listed. A lite template is available for power users.

Highlighted Details

• Benchmark Performance: In tests with Claude Sonnet 4.6, nopua found +104% more hidden bugs, increased proactive initiative (+355%), and improved investigation depth (+83%) compared to a standard agent.
• Multi-Language Support: Offers distinct prompts and configurations for 7 languages, exceeding competing skills.
• Philosophical Foundation: Integrates principles from the Dao De Jing, emphasizing intrinsic motivation, selflessness, and "effortless action" (Wu Wei) over extrinsic pressure.

Maintenance & Community

The project is authored by WUJI (wuji-labs). Community engagement appears centered around WeChat groups and direct contact with the author, with no other public community channels (like Discord or Slack) explicitly mentioned.

Licensing & Compatibility

Released under the permissive MIT License, allowing for commercial use and integration into closed-source projects. It is built with the OpenClaw ecosystem in mind.

Limitations & Caveats

Using nopua concurrently with fear-based "PUA" skills is discouraged due to conflicting motivational frameworks. While the skill file is small, power users with extensive existing agent configurations might find the full 29KB redundant or prefer to extract only the core philosophical components ("Dao") using the provided lite template. Manual installation across different agent frameworks may present a barrier.