AI Coding Playbook

Overview

The AI Coding Playbook is a free, open-source guide to working with agentic coding tools on real codebases. It was written from months of running Claude Code and Gemini CLI against production repos, and it captures the handful of ideas and patterns that survived that work. The book takes a concepts-first, tool-specific-second approach so the material stays useful as the vendor tools evolve.

What’s inside

• 21 chapters covering context economics, briefing vs. prompting, memory hierarchies, slash commands, skills, MCP servers, hooks, subagents, research workflows, test generation, UI generation, local models, and agent SDKs.
• Drop-in assets: 4 slash commands and 2 skills usable across both Claude Code and Gemini CLI, 5 hook scripts for policy enforcement and automation, configuration templates for memory files and MCP setups, and starter ~/.claude/ and ~/.gemini/ structures you can copy directly.
• Six end-to-end playbooks: triaging vague bug reports without jumping to a fix, multi-file migrations with subagent fan-out gated by pattern approval, feature development from spec to shipped with plan-before-edit, read-only on-call triage, research-driven decisions producing committable ADRs, and 90-second PR triage across large queues.

Who it’s for

Solo developers shipping SaaS products, engineering teams rolling out AI coding practices across a codebase, and anyone who wants to understand agentic coding fundamentals beyond marketing demos.

Why it exists

Most agentic-coding material in 2026 is either vendor-specific demo content or surface-level listicles. The playbook focuses on the disciplines that actually make agents useful on real codebases — context discipline, briefing structure, the four primitives (slash commands, skills, MCP, hooks), and the failure modes that show up only after a few weeks of daily use.

License

MIT. Fork it, remix it, build a course on top of it, translate it — just don’t claim you wrote it.