Back to roadmap
EnglishAbout 10 min read

resources/ Index

resources/README.en.md

resources/ Index

This is the repo's reference area: supplementary material that sits outside the main path and is meant to be opened when needed. Each file has a distinct role.


#7 References + When to Read Each

FileRoleWhen to ReadLines
glossary.en.md30-second term lookupYou hit terms like LLM / RAG / token / agent / vector DB / streaming / batch API while reading a stage~210
cli-agents-guide.en.md8 CLI agents comparedFirst time choosing among Claude Code / Codex / OpenCode / Gemini CLI / goose / Aider / Hermes Agent / Grok Build~134
mcp-skills-catalog.en.md81+ integration catalogYou want Claude Code connected to Notion / Obsidian / Excel / Postgres / Slack / other real tools~775
schema-design-cheatsheet.en.md5 function-schema rules + 5 anti-patternsYou are writing a tool schema / MCP server schema / function calling and the LLM picks the wrong tool or arguments~159
cookbook.en.md6 step-by-step recipesYou want to build a first Skill / MCP server / Office integration / NotebookLM flow / Zotero flow / local LLM in 30-50 minutes~620
setup-guide.en.mdFrom-zero setup guideNo dev background; first time creating an API key, installing Python, or using Claude Code~400
style-guide.en.mdFormat and wording rules before PRsYou want to contribute to the repo, add entries, or improve translations~338

Together these are about ~2500 lines of reference. That sounds large, but each file is read at a different moment. You do not read all of them at once; you open the relevant one for 30 seconds to 45 minutes.


#Entry Points: "What Am I Trying to Do?"

#🆕 I Have Never Written Code / This Is My First AI Agent Setup

setup-guide.en.md (30-45 minutes from zero)

#🆕 I Am Just Starting to Learn AI Agents

You do not need any reference first. Start with the main READMEStage 0. When a term is unclear, come back to glossary.en.md.

#🛠 I Need to Choose a CLI Agent

cli-agents-guide.en.md (CLI comparison + recommendations by use case)

#🔌 I Want to Connect Claude Code to Tool X (Notion / Excel / Postgres / etc.)

mcp-skills-catalog.en.md (81+ integrations in 16 categories)

#🍳 I Want to Build My First Skill / MCP Server / Word Integration

cookbook.en.md (6 step-by-step recipes)

#📐 I Wrote a Tool Schema and the LLM Is Not Following It

schema-design-cheatsheet.en.md (5 rules + 5 anti-patterns)

#📚 I Hit an Unclear Term While Reading a Stage

glossary.en.md (30-80 words per term + which stage goes deeper)

#🤝 I Want to Send a PR / Translate / Add a New Entry

style-guide.en.md + ../CONTRIBUTING.en.md


#Duplication?

Duplication is deliberately avoided. Here is how each reference relates to the main path and to the other references:

  • glossary is a 30-second lookup, stage text is a 3-5 minute read, and cookbook is a 30-50 minute build — three depths, no overlap.
  • schema-design-cheatsheet overlaps with cookbook 2 (building an MCP server), but the cheatsheet explains the rules for writing a schema while the cookbook explains how to get a server running. Which one you want depends on where you are stuck.
  • cli-agents-guide is a reference table; the catalog is plug-in tools — two different levels.
  • Relationship to Hello-Agents: Hello-Agents is the most complete Chinese-language agent course available, and it goes deep. This repo works from a different angle, roadmap + curated catalog + hands-on recipes, and does not replace it. Stage 5.3 and cookbook 1 both cross-reference Hello-Agents Extra08, "How to Write a Good Skill", as deeper follow-up reading.

#Language Coverage

Filezh-TW (canonical)zh-HansEnglish
glossary
cli-agents-guide
mcp-skills-catalog
schema-design-cheatsheet
cookbook
setup-guide
style-guide

#Standards for Adding a New Reference

A new reference file should not be added casually. It must:

  1. Avoid duplicating any existing role in the table above.
  2. Solve a problem the main path cannot cover well. If 50 lines in Stage X would cover it, put it in that stage.
  3. Be expected to receive 3+ cross-references from stages or branches. If it only serves one stage, keep it in that stage.

Possible future references:

  • cost-calculator-guide.md: cross-provider pricing. Stage 1 covers enough for now.
  • troubleshooting-guide.md: common error runbook. Existing material is enough until more community reports arrive.
  • prompt-patterns-guide.md: CoT / few-shot template library. Stage 2 already covers the basics; a deeper version can wait for community PRs.

If the community wants one of these added, open an issue to discuss it.