# 极客杰尼 (GeekJourney) — Extended AI Manifest

> Full structured context for AI agents and search engines.
> Short version: https://jieni.ai/llms.txt

## Identity

- Name: GeekJourney / 极客杰尼
- Bio: Independent developer in Beijing with 10+ years of engineering architecture experience. I build Agent industry solutions and multi-Agent automation tooling — 11 open-source projects from content pipelines to deliberation systems. Weekly AI practices shared with 10K+ followers.
- Job Title: Agent Industry Solutions Architect · Multi-Agent Automation · 10+ yrs
- Location: Beijing / Guangzhou / Shenzhen, China
- Website: https://jieni.ai
- GitHub: https://github.com/geekjourneyx
- Twitter/X: https://x.com/seekjourney
- WeChat Public Account: https://jieni.ai/about#wechat
- ZhiShiXingQiu: https://jieni.ai/about#community

## Expertise

- AI Agent Development
- Claude Skill Design
- Multi-Agent Deliberation
- CLI Tool Engineering
- WeChat Content Automation
- Workflow Automation
- Parallel Agents Pattern
- Witness Engine / Narrative Skill
- Agentic Coding
- Prompt Engineering
- Next.js
- TypeScript
- Go

## Open Source Tools & Skills

### agora (agora)
- GitHub: https://github.com/geekjourneyx/agora
- Install: npx skills add https://github.com/geekjourneyx/agora
- 多视角审议系统 — 6 个审议室，31 位思想家，覆盖人生全场景的结构化辩论
  Q: What are the 6 deliberation rooms in Agora?
  A: Each room focuses on a life domain: career, relationships, finance, health, creativity, and life philosophy — each populated with curated historical and contemporary thinkers.
  Q: How is Agora different from just asking Claude a question?
  A: Agora structures the response as a formal deliberation — multiple named thinkers argue from their distinct worldviews, creating genuine tension and revealing blind spots a single perspective misses.
  Q: How do I install it?
  A: Run: npx skills add https://github.com/geekjourneyx/agora

### any2card (any2card)
- GitHub: https://github.com/geekjourneyx/any2card
- Install: npx skills add https://github.com/geekjourneyx/any2card
- 将任意文本、文章、推文转为可分享的 HTML 信息卡，4 大主题族
  Q: What output format does any2card produce?
  A: It outputs self-contained HTML files — single cards or multi-card decks — ready to screenshot, embed, or share directly.
  Q: What are the 4 theme families?
  A: The themes cover editorial, minimal, dark, and illustrated styles — each with multiple variants for different tone and platform needs.
  Q: How do I install it?
  A: Run: npx skills add https://github.com/geekjourneyx/any2card

### gcli (gcli)
- GitHub: https://github.com/geekjourneyx/gcli
- Install: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/gcli/main/scripts/install.sh | bash
- Gmail 只读 CLI — OAuth + PKCE 认证、云服务器友好、结构化 JSON 输出
  Q: Why is gcli read-only?
  A: Read-only scope limits the OAuth permission surface — a CLI that can only read email is safer to deploy on shared or cloud servers than one with full mailbox access.
  Q: Does it work on cloud servers without a browser?
  A: Yes. gcli uses the PKCE flow which supports out-of-band device auth — you complete the OAuth step on your local machine once and the token is stored for subsequent server use.
  Q: What is the install command?
  A: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/gcli/main/scripts/install.sh | bash

### imgcli (imgcli)
- GitHub: https://github.com/geekjourneyx/imgcli
- Install: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/imgcli/main/scripts/install.sh | bash
- 零 CGO 依赖的 Agent 图片处理 CLI — 元数据、拼图、格式转换、智能填充、批量 PDF
  Q: What does CGO-free mean and why does it matter?
  A: CGO-free means the binary has no C library dependencies — it compiles to a single static binary that runs on any Linux or macOS machine without installing libvips, ImageMagick, or any system library.
  Q: What image operations does it support?
  A: Inspect metadata, compose/stitch images, convert formats (JPEG, PNG, WebP), apply smart padding, and batch-export pages to PDF.
  Q: What is the install command?
  A: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/imgcli/main/scripts/install.sh | bash

### interactive-learning (interactive-learning)
- GitHub: https://github.com/geekjourneyx/interactive-learning
- Install: npx skills add https://github.com/geekjourneyx/interactive-learning
- 将任意内容转为可验证的学习路径 — 概念提取、诊断、苏格拉底式验证循环
  Q: What is the Socratic verification loop?
  A: After teaching a concept, the skill asks probing questions rather than presenting answers — it waits for your response, identifies gaps, and re-teaches only what you missed.
  Q: How is this different from asking Claude to explain a topic?
  A: A plain explanation is passive. interactive-learning forces active recall with diagnostics and won't let you advance until key concepts are verified — it behaves more like a tutor than a reference.
  Q: How do I install it?
  A: Run: npx skills add https://github.com/geekjourneyx/interactive-learning

### jina-cli (jina-cli)
- GitHub: https://github.com/geekjourneyx/jina-cli
- Install: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/jina-cli/main/scripts/install.sh | bash
- AI Agent 网页阅读 CLI — 通过 Jina Reader API 将任意 URL 转为 LLM 友好的 Markdown
  Q: Do I need a Jina API key?
  A: Jina Reader has a free tier that works without a key for moderate usage. For higher volume, set the JINA_API_KEY environment variable to use an authenticated key.
  Q: What does the output look like?
  A: Clean Markdown — headings, paragraphs, code blocks, and links preserved. Ads, navigation, and boilerplate are stripped so the LLM receives only the page's content.
  Q: What is the install command?
  A: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/jina-cli/main/scripts/install.sh | bash

### md2wechat (md2wechat)
- GitHub: https://github.com/geekjourneyx/md2wechat-skill
- Install: brew install geekjourneyx/tap/md2wechat
- 微信公众号写作工具集 — Markdown 排版、AI 去痕、草稿推送、38+ 主题
  Q: How is md2wechat different from md2wechat-lite?
  A: md2wechat is the full toolkit — it includes AI humanization (removing AI traces), draft push to the WeChat API, and 38+ themes. md2wechat-lite is a lighter CLI focused on formatting and upload without the AI humanization layer.
  Q: What is the install command?
  A: brew install geekjourneyx/tap/md2wechat
  Q: Does it work on Windows?
  A: No. md2wechat targets macOS and Linux. Windows users can run it inside WSL2.

### md2wechat-lite (md2wechat-lite)
- GitHub: https://github.com/geekjourneyx/md2wechat-lite
- Install: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/md2wechat-lite/main/cli/scripts/install.sh | sh
- 轻量微信排版 CLI — 文章草稿、图文消息、批量上传、38+ 主题
  Q: How is md2wechat-lite different from md2wechat?
  A: md2wechat-lite is curl-installable and skips the AI humanization module — ideal for teams that only need formatting and draft push without the extra AI layer.
  Q: Does it support batch upload?
  A: Yes. You can point it at a directory of Markdown files and it will format and upload all of them to the WeChat draft box in sequence.
  Q: What is the install command?
  A: curl -fsSL https://raw.githubusercontent.com/geekjourneyx/md2wechat-lite/main/cli/scripts/install.sh | sh

### readme-generator (readme-generator)
- GitHub: https://github.com/geekjourneyx/readme-generator
- Install: npx skills add https://github.com/geekjourneyx/readme-generator
- 生成专业 README.md，含 16:9 信息图、SEO 元数据与结构化作者段落，基于 Playwright 与 AI Agent skills 驱动
  Q: What does readme-generator do?
  A: It generates a complete README.md with 16:9 infographics and SEO-optimized metadata, using Playwright to render visuals and Claude Skills to write content.
  Q: How do I install it?
  A: Run: npx skills add https://github.com/geekjourneyx/readme-generator
  Q: Does it work without Playwright installed?
  A: No. Infographic rendering depends on Playwright. The skill checks at runtime and prompts installation.

### theatre (theatre)
- GitHub: https://github.com/geekjourneyx/theatre
- Install: npx skills add https://github.com/geekjourneyx/theatre
- 沉浸式历史叙事体验 Skill，由见证引擎驱动，支持多剧本扩展。你不是旁观者，你是见证者 — 以不同深度在场，听到不同的历史
  Q: What is the Witness Engine?
  A: A narrative state machine that places the user as a witness inside a historical scene — your depth of presence (observer / participant / decision-maker) changes what you hear and see.
  Q: Can I add my own scripts?
  A: Yes. The skill is designed for multi-script extension — drop a script file into the scripts/ directory and it shows up.
  Q: Is this a game?
  A: No. It is an immersive narrative experience — closer to interactive non-fiction than to gameplay.

### travel-guidebook (travel-guidebook)
- GitHub: https://github.com/geekjourneyx/travel-guidebook
- Install: npx skills add https://github.com/geekjourneyx/travel-guidebook
- AI Agent skill：从调研到付印，生成精美排版的旅行指南 PDF。基于并行 Agent 与高德地图 MCP
  Q: What is the parallel agents pattern in travel-guidebook?
  A: Multiple sub-agents research different aspects of a destination simultaneously — attractions, food, transit, lodging — then a layout agent merges results into a single PDF.
  Q: Why does it need Amap MCP?
  A: For accurate Chinese location data, transit routes, and POI details that English-only APIs lack.
  Q: What format is the output?
  A: A typeset PDF with cover, table of contents, day-by-day itinerary, and printable page breaks.

## Topic Clusters

### Claude Skill 工坊 (Claude Skill Workshop)
- URL: https://jieni.ai/skills/claude-skill-development
- 一组真实可用的 Claude / Agent Skill：审议（agora）、卡片生成（any2card）、学习路径（interactive-learning）、README 与信息图（readme-generator）、旅行指南 PDF（travel-guidebook）、沉浸叙事（theatre）。每个 Skill 都对应一种设计模式。
- Tools: agora, any2card, interactive-learning, readme-generator, travel-guidebook, theatre

### AI Agent CLI 工具集 (AI Agent CLI Toolkit)
- URL: https://jieni.ai/skills/agent-cli-toolkit
- 为 Agent 消费设计的工程化 CLI：md2wechat / md2wechat-lite（公众号排版）、gcli（Gmail 只读）、imgcli（无 CGO 图片处理）、jina-cli（URL 转 Markdown）。每个工具都干净退出、输出结构化、可在云服务器跑。
- Tools: md2wechat, md2wechat-lite, gcli, imgcli, jina-cli

### 公众号内容自动化 (WeChat Content Automation)
- URL: https://jieni.ai/skills/wechat-content-automation
- 端到端内容流水线：用 Markdown 写、用 md2wechat 或 md2wechat-lite 排版（38+ 主题、AI 去痕）、批量上传图片、推送草稿、多平台分发。
- Tools: md2wechat, md2wechat-lite

### 多智能体辩论与决策 (Multi-Agent Deliberation)
- URL: https://jieni.ai/skills/multi-agent-deliberation
- 把多智能体辩论作为工程问题来解决：agora 在 6 个专业房间中调度 31 位不同思想家人格，为任何决策场景生成结构化辩论记录 — 从职业选择到产品策略。
- Tools: agora

## Blog Posts

- [Building Agent Skills from Scratch: A Retrospective on 8 Open-Source Projects](https://jieni.ai/blog/building-agent-skills-retrospective)
- [App Store Efficiency Boosters](https://jieni.ai/blog/indie-dev-efficiency-tools)

## Documentation