Playbook
No missions available in this destination.
Establishing uplink...
Drag to orbit / Scroll to zoom / Right-drag to pan / Select a world or a satellite
— THE AI DEV PLAYBOOK —
A structured map of GitHub Copilot tools and concepts across the SDLC: planning, development, review, testing, security, and operations. Note: this is a personal learning resource and some details may be outdated — always refer to the latest GitHub documentation ↗ for the source of truth.
Introduction
はじめに 3 items
👤 Mona
A platform-level overview of GitHub and GitHub Copilot.
-
NO.01GitHub
The AI-native developer platform used by more than 240 million developers. From Pull Requests to GitHub Copilot, GitHub continues to shape the history of software development.
Read details -
NO.02GitHub Copilot
The world's most widely used AI development tool. An orchestrator that lets you freely combine major AI models with a variety of surfaces.
Read details -
NO.03Hands-on
A workshop where you get hands-on with the concepts in this playbook by rebuilding a simplified version of this very playbook site from scratch — together with Copilot. Touches MCP / Instructions / Agent Skills / Plan mode / Cloud Agent / Code Review / Agentic Workflow end-to-end.
Read details
Plan
企画 8 items
👤 Team Manager
Plan with Issues and Projects, then prepare the AI harness with MCP servers, instruction files, agent skills, and custom agents.
-
NO.3.4Collaboration Tools
GitHub's toolbox for working together. Starting from the Pull Request GitHub invented, a quick tour of Issues & Sub-issues, Discussions, and Projects, with the "why it's good" for each and the official links to dig deeper.
Read details -
NO.3.5Harness Engineering
The discipline of designing the scaffold that lets agents operate safely and effectively — tools, constraints, permissions, and external connections.
Read details -
NO.04MCP Server
MCP Server (Model Context Protocol)
MCP is a method for providing additional context to AI models.
Read details -
NO.05Instructions
Persistent rule files that give AI standing development guidelines at the repository or file level, so every teammate's Copilot follows the same conventions.
Read details -
NO.06Agent Skills
Reusable instruction packs that teach Copilot agents how to handle specialized tasks — loaded automatically when the request matches, no re-explaining required.
Read details -
NO.07Custom Agent
A specialist profile that bundles role, tools, and behavior together for Copilot. Switch between Planner, Reviewer, and Tester personas with their own permissions.
Read details -
NO.7.5Hooks
A mechanism for injecting custom scripts into the agent execution lifecycle (session start, prompt submission, pre/post tool execution, error, session end). Configured via `.github/hooks/hooks.json`. The most powerful hook is PreToolUse, which can deny specific commands (rm -rf, sudo, pushes to production, etc.) to block the agent. Works with both Copilot CLI and Cloud Agent.
Read details -
NO.8.5Context Engineering
The art of designing the context you pass to AI — not too little, not too much, just the information needed to complete the task.
Read details
Develop
開発 8 items
👤 Junior Developer
Code in shared Codespaces environments with Plan / Agent mode, Copilot Chat, and the CLI. Delegate work to Cloud Agent for parallel development.
-
NO.3.5Copilot Chat
The synchronous conversation surface for Copilot. Available across IDE, GitHub.com, and Mobile, with the deepest harness on the IDE side.
Read details -
NO.14Copilot CLI
Copilot living in the terminal. An IDE-agnostic "colleague in the terminal" that lets you move between planning → implementation → review with just Shift+Tab.
Read details -
NO.15Cloud Agent
A "doppelganger" that works through the night even when you leave the IDE. Assign an Issue and it implements, verifies, and returns with a PR from a GitHub Actions runner.
Read details -
NO.15.5Partner agents
A mechanism to launch coding agents other than Copilot — such as Claude or Codex — from VS Code and GitHub with the same feel as Copilot.
Read details - NO.16.7
Copilot App
A purpose-built desktop app for agent-driven development. Run multiple agents in parallel and handle issue triage, implementation, review, and merge in one place—without bouncing between IDE, terminal, and browser.
Read details - 🪟NO.16.8
VS Code Agents Window
A dedicated, chat-first VS Code window that spans every workspace at once. Start, monitor, and review agent sessions across all your projects from a single sessions list, instead of opening one window per repository.
Read details -
NO.17Codespaces
Customizable remote development environments hosted in the cloud. The ideal AI-driven development playground for safely sandboxing AI experiments and work.
Read details -
NO.17.5Spaces
A shared place to organize the context you want Copilot to use: code, issues, pull requests, docs, notes, images, uploads, and instructions that ground Copilot's answers.
Read details
Review
レビュー 2 items
👤 Senior Developer
Let Copilot perform automated code review so humans can focus on final judgment.
- 🔀NO.8.8
Pull Requests
The mechanism GitHub created in 2008 that changed how developers collaborate. The propose → discuss → review → merge flow, now backed by rulesets, automated checks, and Copilot review.
Read details -
NO.09Copilot Code Review
Copilot can be assigned as a reviewer — manually or automatically — to read code intent and return inline comments, PR summaries, and fix suggestions. Goes from pre-review noise reduction all the way to full automation.
Read details
Test & Secure
テスト & 品質保証 6 items
👤 DevSecOps Engineer
Automate tests with GitHub Actions, maintain quality with Code Quality, and secure code with Code Scanning, Secret Scanning, and Dependabot.
-
NO.19.2GitHub Actions
GitHub's native CI/CD and automation platform. Steps written in .github/workflows/*.yml run automatically on push / PR / schedule. Public repos are free; private repos get a free tier plus pay-as-you-go. Cloud Agent and Copilot Code Review also run on top of this.
Read details -
NO.19.3GHAS
GitHub Advanced Security
GitHub's paid security suite, sold as two products. Secret Protection ($19) keeps credentials out of the repo, Code Security ($30) finds vulnerabilities in the code. Billed per active committer. Public repos are free.
Read details -
NO.19.4Secret Scanning
GitHub's secret detection feature that automatically finds API keys and tokens in commits, issues, PRs, and history. Core coverage is free for public repositories; private and internal repositories require Secret Protection.
Read details -
NO.19.5Code Scanning
Code scanning finds vulnerabilities with CodeQL static analysis (SAST). Eligible findings can receive Copilot Autofix suggestions or be assigned to Copilot for remediation.
Read details - 🩺NO.19.6
Code Quality
GitHub Code Quality catches reliability and maintainability issues before merge using CodeQL and AI, suggests fixes, reports coverage and quality scores, and enforces quality gates with rulesets.
Read details -
NO.19.7Dependabot
GitHub's official bot that monitors repository dependencies for vulnerabilities and opens automated fix PRs. Built on the dependency graph, it runs in three layers — Security alerts, Security updates, and Version updates. Free for public and private repos.
Read details
Operate
運用 5 items
👤 Whole Team
Automate operations with agentic workflows, visualize AI usage with Copilot Metrics, and accumulate knowledge through Memory.
-
NO.16Agentic Workflow
A new kind of automation that runs AI agents on top of GitHub Actions. Write what you want done in natural language Markdown and get workflows that can reason and act.
Read details -
NO.16.5Token Optimization
Tokens get cheaper by raising agent quality, not by shaving prompts. Use model choice, context engineering, deterministic guardrails, and the right agent configs to make every token count.
Read details -
NO.18Copilot Metrics
Measure Copilot adoption, engagement, code generation, and delivery impact through dashboards and exportable APIs.
Read details -
NO.19Copilot Memory
A mechanism for Copilot to remember and reuse knowledge learned while working in a repository, passing it on to future selves and the whole team. Gets smarter the more you use it.
Read details - 🛠️NO.19.5
Operational Tools
GitHub's toolbox for putting collaboration into operation. A quick tour of private Packages and the container registry, and Integrations (GitHub Apps).
Read details
Administration
管理 7 items
👤 Enterprise Admin
Govern licenses, policies, and billing at the Enterprise and Organization level. Decide safely who gets access to what.
- 🌏NO.29.2
EMU & Data Residency
Compare the three GitHub Enterprise Cloud setups (personal accounts, EMU, and data residency) with self-hosted GHES. Covers EMU user management and restrictions, how GHE.com keeps your data in a region such as Japan, features unavailable on GHE.com and GHES, and how Copilot fits in.
Read details -
NO.29.5Enterprise Setup
From creating a GitHub Enterprise Cloud (EMU) account to configuring SSO — spin up a 30-day free trial, set up OIDC or SAML via Entra ID, and get your team ready to log in.
Read details -
NO.30GHE & Copilot Licences
Copilot billing, assignment, multi-org policies, and external collaborators. GHE Cloud licence consumption, unaffiliated users, Visual Studio bundles, and budget alerts.
Read details -
NO.30.5Usage-based Billing
Usage-based Billing (UBB)
On June 1, 2026 GitHub Copilot moved to usage-based billing (UBB). Cost is tied to the tokens actually processed (input, output, cached), and usage is shown and budgeted in GitHub AI credits (1 credit = $0.01 USD). The core principles, aligned with the docs.
Read details - ⚖️NO.30.8
Governance
Control who can do what. Organization structure, teams, access roles, policies, rulesets, and Copilot managed settings.
Read details - 💰NO.30.9
Cost centers & budgets
Group metered spending with cost centers, then manage budgets, alerts, usage limits, and included AI credit pools.
Read details - 🚚NO.31
Migrations
Compare non-EMU ↔ EMU and GHES → GHE.com (data residency) migrations. When to use GEI or ELM, prerequisites, data that is not migrated, limits, and follow-up tasks, in a foldable table.
Read details
