Projects View on GitHub

AI Best Practices -- A Practical Guide from Anthropic Courses

Language: English | Turkish

This repo is a distilled summary of 10 courses from Anthropic's Skilljar training platform. The goal is not to pile up theory; it is to give someone who wants to use AI better than average, without spending hours watching videos a consistent, practical reference they can keep close.

What Is This Repo Based On?

The concepts were extracted from notes across these 10 courses:

CourseWhat It Teaches
AI Capabilities and LimitationsHow models work (next-token prediction), what they cannot do (hallucination, knowledge cutoff, working memory), and how to diagnose which mechanism caused a mistake.
AI Fluency Framework and FoundationsThe 4D Framework: Delegation, Description, Discernment, Diligence. A universal frame for responsible, effective AI work.
Claude 101Claude's core capabilities: Projects, Artifacts, Connectors, Research mode, custom instructions, memory.
Claude Code 101The agentic loop in the terminal: explore -> plan -> code -> commit. CLAUDE.md, context management, plan mode.
Claude Code in ActionClaude Code in real projects: MCP server setup, UI testing with Playwright, GitHub Actions, hooks, SDK.
Introduction to Claude CoworkCowork as a delegation environment that handles files, scheduled tasks, and plugins.
Introduction to agent skillsSkills: reusable, automatically triggered instruction packages built around SKILL.md.
Introduction to subagentsSubagents: specialized helpers that run in a separate context without polluting the main chat.
Introduction to Model Context ProtocolMCP fundamentals: tools, resources, prompts, and client-server architecture.
Model Context Protocol Advanced TopicsSampling, logging, transport (stdio / HTTP / streamable), roots, progress notifications.

Language Structure

The main repository structure is English. Turkish is kept as an optional mirrored version under tr/:

How To Use This Repo

There are three layers:

  1. Skimmable: how-to-write-better.md -- all concepts in one file, with real-world examples, without going too deep. Read it over a tea break.
  2. Detailed: concepts/ -- focused documents for each concept, deep enough without drowning you.
  3. Applied:
  • checker/ -- a drop-in folder that audits an existing project against AI best practices.
  • starter-template/ -- a fill-in template for starting a new project from zero.

Navigator

Quick Start

Concept Deep Dives

Applied


This repo is designed as a living document. As the concepts change, the related files should be updated. Each file starts with a short "what this is for / when to read it" section.