AI Product Research Skill for Codex

一个面向 Codex 的深度 AI 产品调研技能。它把公开来源、GitHub 数据、产品定位、用户路径、市场趋势、PM 分层学习和 HTML 报告输出组织成一套可复用工作流。

A deep AI product research skill for Codex. It turns public sources, GitHub signals, product positioning, user journeys, market trends, PM-level reasoning, and HTML reporting into a reusable workflow.

为什么值得 star / Why star this repo?

• 可复用的产品研究框架：适合研究 AI 产品、GitHub 项目、SaaS、桌面应用、开源工具、模型平台和产品趋势。

• 面向 PM 的学习结构：不仅描述功能，还训练从产品负责人、Head of Product、Staff PM、Senior PM 到初级 PM 的分层判断。

• 默认生成可读 HTML 报告：报告结构包含沉浸式 hero、指标面板、功能树、用户旅程、竞品矩阵、趋势分析和 scorecard。

• 强制区分事实与推断：要求优先使用官方来源，并标注估算、假设和证据边界。

• 可扩展到视频/音频：研究报告可作为后续视频脚本、音频讲解或 storyboard 的事实基础。

• Reusable product research framework: Useful for AI products, GitHub repositories, SaaS tools, desktop apps, open-source projects, model platforms, and product trend studies.

• PM-level learning structure: Goes beyond feature lists and teaches reasoning from product leader to beginner PM levels.

• HTML report by default: Produces rich local reports with hero sections, metric panels, feature trees, journeys, competitor matrices, trend analysis, and scorecards.

• Fact vs inference discipline: Prefers primary sources and labels estimates, assumptions, and evidence limits.

• Ready for video/audio follow-up: The report can become the source of truth for scripts, narrated explainers, and storyboards.

适用场景 / Use Cases

中文

• 调研一个 AI 产品的定位、目标用户、核心工作流和商业机会。
• 分析 GitHub 开源 AI 项目的 stars、forks、release cadence、issue/PR 活跃度和产品成熟度。
• 为产品经理生成可学习的 PM 分层分析，而不只是摘要。
• 产出可浏览的本地 HTML 产品研究报告。
• 在制作视频或音频讲解前，先生成事实可靠的研究底稿。

English

• Research the positioning, target users, workflows, and business opportunities of an AI product.
• Analyze open-source AI repositories using stars, forks, release cadence, issue/PR activity, and product maturity signals.
• Generate PM-level learning material instead of shallow summaries.
• Produce a readable local HTML product research report.
• Prepare a source-backed research base before creating video or audio explainers.

包含内容 / What is included

.
├── SKILL.md
├── agents/
│   └── openai.yaml
├── references/
│   ├── capability-check.md
│   ├── cross-skill-handoff.md
│   ├── data-estimation-playbook.md
│   ├── pm-level-framework.md
│   ├── report-quality-gate.md
│   ├── report-template.md
│   ├── trend-analysis-framework.md
│   ├── ui-function-tree-guide.md
│   ├── video-learning-output-guide.md
│   ├── video-production-brief-template.md
│   ├── video-qa-gate.md
│   └── video-script-quality-gate.md
├── CONTRIBUTING.md
├── SECURITY.md
└── LICENSE

安装 / Installation

中文

将本仓库克隆到你的 Codex skills 目录，目录名保持为 ai-product-research 或在 Codex 中按你的技能加载方式注册：

git clone https://github.com/hexiao160705-dev/ai-product-research-skill.git "$env:USERPROFILE\.codex\skills\ai-product-research"

然后在 Codex 中使用：

使用 $ai-product-research 调研 https://github.com/openai/openai-cookbook

English

Clone this repository into your Codex skills directory. Keep the directory name as ai-product-research, or register it according to your Codex skill loading setup:

git clone https://github.com/hexiao160705-dev/ai-product-research-skill.git "$env:USERPROFILE\.codex\skills\ai-product-research"

Then use it in Codex:

Use $ai-product-research to research https://github.com/openai/openai-cookbook

示例 Prompt / Example Prompts

GitHub 项目调研 / GitHub repository research

使用 $ai-product-research 调研这个项目：https://github.com/browser-use/browser-use
请输出中文 HTML 报告，重点看产品定位、用户场景、活跃度、竞争格局和 PM 学习价值。

Use $ai-product-research to research https://github.com/langchain-ai/langchain.
Focus on positioning, developer workflow, release momentum, ecosystem strategy, and PM-level lessons.

SaaS 产品分析 / SaaS product analysis

使用 $ai-product-research 调研 Cursor。请比较它和 Windsurf、GitHub Copilot 的定位差异，并输出 HTML 报告。

Use $ai-product-research to analyze Perplexity as a product.
Include positioning, target users, first-use journey, retention assets, monetization, and market trend insights.

AI 工具竞品研究 / AI tool competitor research

使用 $ai-product-research 对比 3 个 AI agent builder 产品，输出产品经理视角的机会判断。

Use $ai-product-research to compare three AI video generation tools.
Separate facts from assumptions and include a competitor/category position matrix.

视频脚本前置研究 / Research before video script

使用 $ai-product-research 先调研这个 AI 产品并生成 HTML 报告。
报告确认后，我再要求你基于同一事实来源生成视频脚本。

Use $ai-product-research to create the source-backed research report first.
Do not generate a video script until I approve the report.

输出物 / Outputs

默认输出是本地 HTML 报告：

projects/<product-slug>/report.html

可选输出包括：

• projects/<product-slug>/video-script.md
• projects/<product-slug>/audio-script.md
• projects/<product-slug>/video-production-brief.md

The default output is a local HTML report. Optional outputs include video scripts, audio scripts, and video production briefs when explicitly requested.

研究原则 / Research Principles

• 优先使用官方来源：官网、GitHub README、releases、docs、pricing、official blog、package registry、app store 页面。

• 对现代产品、GitHub 数据、release、价格、指标和公司信息进行实时检索。

• 区分事实、证据支持的解读和推断。

• 不编造精确数据；缺少内部数据时使用公开代理指标并说明置信度。

• 默认不写标准技术架构章节，除非用户明确要求代码或实现分析。

• 以产品判断为目标，而不是堆砌资料。

• Prefer primary sources: official sites, GitHub README, releases, docs, pricing, official blogs, registries, and app store pages.

• Browse current sources for modern products, GitHub activity, releases, pricing, metrics, and company information.

• Separate facts, evidence-backed interpretation, and inference.

• Do not invent exact metrics; use public proxies and label confidence when internal data is unavailable.

• Do not include a standard technical architecture section unless requested.

• Optimize for product judgment, not raw information volume.

能力边界 / Boundaries

• 这个 skill 不会自己生成视频、音频或 TTS；它会在用户明确要求后准备脚本或交接 brief。

• 它不是投资建议工具，也不替代安全审计、法律审查或财务尽调。

• GitHub stars、forks、issues、downloads 等都是代理指标，不等于真实用户规模或收入。

• This skill does not generate video, audio, or TTS by itself; it prepares scripts or handoff briefs only when explicitly requested.

• It is not investment advice and does not replace security audits, legal review, or financial due diligence.

• GitHub stars, forks, issues, and downloads are proxy signals, not exact user or revenue metrics.

Roadmap

• Add more reusable example reports.
• Add optional templates for competitor matrices and product scorecards.
• Improve bilingual report patterns.
• Add source-quality scoring guidance.
• Add more examples for GitHub repository, SaaS, desktop app, and AI model platform research.

Contributing

欢迎提交 PR 改进研究框架、报告模板、质量门槛和示例。请先阅读 CONTRIBUTING.md 和 SECURITY.md，不要提交任何 API key、客户数据、内部业务资料或本地生成报告。

Contributions are welcome. Please read CONTRIBUTING.md and SECURITY.md first. Do not submit API keys, customer data, internal business material, or local generated reports.

License

MIT License. See LICENSE.