[Skill] A Chinese-first product decision agent distilled from On Practice and On Contradiction

[atdy]

I built Maoxuan Product Agent, an open-source Agent Skill for Chinese-speaking product managers, operators, growth teams, and founders.

Repository: https://github.com/atdy/maoxuan-product-agent

The project is unusual in one specific way: its reasoning engine was designed after complete readings of On Practice, On Contradiction, and related Volume I essays, but it does not behave like a political, historical, or quotation skill. The source method stays in the background.

For a real product question, it tries to:

1. separate facts from assumptions;
2. identify the current bottleneck and product stage;
3. determine which mechanism is driving the result;
4. choose the smallest useful, reversible action;
5. state what not to do yet and which signal should change the decision.

Default answers are in Simplified Chinese and use modern product language. Covered scenarios include requirements prioritization, Roadmaps, DAU and retention declines, growth stagnation, A/B test anomalies, community cold starts, content supply, metric-definition conflicts, executive interruptions, project delays, and cross-team resource conflicts.

Install for Claude Code

npx skills add atdy/maoxuan-product-agent --skill product-decision-agent --agent claude-code -g -y

Then try:

/product-decision-agent 老板临时插了一个积分商城，但这个版本原本在做新手转化，我该怎么处理？

What is included

• a standard SKILL.md package with progressive reference files;
• 36 scenario-based evaluations and deterministic output quality checks;
• 12 standalone, browsable decision cases;
• MIT license, no API key, no network access, and no runtime dependency;
• installation support for Claude Code, Codex, Cursor, and other Agent Skills clients.

Links:

• Website and cases: https://atdy.github.io/maoxuan-product-agent/
• skills.sh listing and security checks: https://www.skills.sh/atdy/maoxuan-product-agent/product-decision-agent
• Evaluation assets: https://github.com/atdy/maoxuan-product-agent/tree/main/evaluation

I would especially value feedback from Chinese product practitioners on trigger accuracy and whether the answer format produces a genuinely useful next decision rather than generic advice.

AI assistance disclosure: Codex was used extensively for research, implementation, testing, packaging, and drafting this post. Claude Code was used for an independent project review. The maintainer reviewed and approved the published project.