References
All technical claims in this workshop are grounded in the primary sources below. Where the product is evolving, prefer the live commands (/help, /model, copilot help <topic>) over any value frozen in a document.
Primary sources — GitHub official documentation
| Document |
What it covers |
| About GitHub Copilot CLI |
Concepts: modes, context management, customization, security, model usage, sandboxes, ACP |
| Using GitHub Copilot CLI |
How-to: sessions, tips, custom instructions/agents/skills, MCP, context commands |
| Best practices for GitHub Copilot CLI |
Models, plan mode, infinite sessions, delegate, workflows, advanced patterns, team guidance |
| Installing GitHub Copilot CLI |
Install paths and prerequisites |
| GitHub Copilot features |
The assistive / agentic / customization / admin taxonomy across surfaces |
| Adding custom instructions for GitHub Copilot CLI |
Instruction file locations and precedence |
| Adding agent skills for GitHub Copilot CLI |
Authoring SKILL.md skills |
| About agent skills |
Skill concepts |
| Creating custom agents |
Agent-profile schema |
| About the rubber duck agent |
Cross-model "second opinion" critique used by /rubber-duck |
Running tasks in parallel with /fleet |
Parallel subagent execution and /tasks monitoring |
| Allowing Copilot CLI to work autonomously (autopilot) |
How autopilot iterates (plan → act → test → verify) until done |
| About GitHub Copilot plugins |
Plugin marketplaces and /plugin management |
| About hooks for GitHub Copilot |
Lifecycle hooks |
| About GitHub Copilot Memory |
Persistent repo memories |
| About Model Context Protocol (MCP) |
MCP concepts |
| Configure MCP servers (JSON structure) |
mcp-config.json schema |
| About GitHub Copilot code review |
Automated code review |
| About cloud and local sandboxes |
Sandboxing model: local vs cloud, session lifecycle, auth, billing |
| Configuring local sandbox settings |
The /sandbox UI: General / Filesystem / Network settings and platform limits |
| Enabling or disabling cloud sandboxes for your organization |
Org policy: Cloud Sandbox access |
| Billing for cloud and local sandboxes |
Compute / Memory / Storage meters for cloud sandboxes |
| Models and pricing for GitHub Copilot |
Premium requests, model costs |
| Copilot CLI command reference |
Full command & flag list |
| Copilot CLI configuration directory |
~/.copilot settings reference |
| Copilot CLI ACP server |
Agent Client Protocol |
| Customization cheat sheet |
When to use which customization feature |
| Plans for GitHub Copilot |
Plan comparison & pricing |
Primary sources — GitHub repositories & packages
Change watchlist
Check these sources before each workshop run. Copilot CLI changes quickly enough that static workshop material should be treated as a snapshot, not a contract.
| Area to watch |
Primary source |
Why it matters |
| CLI releases and command behavior |
github/copilot-cli changelog |
Tracks CLI versions, new slash commands (/settings, /security-review, /worktree), prompt-mode behavior, MCP config loading, hooks, permissions, and bug fixes |
| Copilot-wide product announcements |
GitHub Blog Changelog — Copilot label |
Tracks model releases/deprecations, public previews, plan changes, code review updates, sandboxes, BYOK, and usage metrics |
| Model availability and deprecations |
Supported models, GitHub Blog Changelog — Copilot |
Model names and plan eligibility change frequently; examples should instruct learners to use /model rather than copy fixed names |
| BYOK / external provider models |
Using your own LLM models in GitHub Copilot CLI, Using your LLM provider API keys with Copilot |
Separates client-side BYOK from enterprise-admin configured external-provider models |
| Sandboxes |
About cloud and local sandboxes, GitHub Blog sandbox announcement |
Public-preview behavior, pricing, platform support, and policy controls may change |
| MCP and agent discovery |
MCP management, github/copilot-cli changelog |
MCP config locations, registry behavior, deferTools, and Agent finder are actively evolving |
| Code review behavior |
About GitHub Copilot code review, AGENTS.md support announcement |
Review instructions, draft PR review UX, and timeline behavior affect Demo 2 |
| Usage and billing signals |
Copilot usage metrics API, AI credits usage API announcement |
Workshop owners need to explain AI credit consumption and admin reporting accurately |
Recent changes reflected in this workshop
| Date |
Change |
Source |
| 2026-06-29 |
Claude Opus 4.8 (fast mode) entered preview across Copilot surfaces including Copilot CLI (admin policy off by default) |
GitHub Blog Changelog |
| 2026-06-26 |
MAI-Code-1-Flash reached general availability for Copilot Business and Copilot Enterprise |
GitHub Blog Changelog |
| 2026-06-25 |
Copilot code review now uses the CLI/SDK grep/rg/glob/view file tools (~20% lower review cost) and adds org-level default Medium analysis depth |
GitHub Blog Changelog |
| 2026-06-24 |
Free and Student plans default to Copilot auto model selection only; (Preview) label retired from Microsoft-released models |
GitHub Blog Changelog |
| 2026-06-24 |
/security-review available to all users without --experimental (CLI 1.0.64) |
copilot-cli changelog 1.0.64 |
| 2026-06-23 |
New Copilot CLI terminal interface GA: tabbed Session/Gists/Issues/Pull requests layout, /mcp search GitHub MCP Registry install, and /theme color modes |
GitHub Blog Changelog |
| 2026-06-19 |
ai_credits_used added to user-level Copilot usage metrics API reports |
GitHub Blog Changelog |
| 2026-06-18 |
Generated release notes now credit the developer alongside @copilot for Copilot cloud-agent PRs |
GitHub Blog Changelog |
| 2026-06-18 |
Copilot code review reads repository-root AGENTS.md |
GitHub Blog Changelog |
| 2026-06-17 |
Enterprise-admin configured external-provider models appear in Copilot CLI /model |
GitHub Blog Changelog |
| 2026-06-11 |
/settings provides a unified, schema-driven settings UI and inline setting changes |
GitHub Blog Changelog |
| 2026-06-10 |
/security-review added as an experimental public-preview CLI command |
GitHub Blog Changelog |
| 2026-06-04 |
1M-token context windows and configurable reasoning levels available on supported models |
GitHub Blog Changelog |
| 2026-06-02 |
Rubber duck and voice input generally available; prompt scheduling and new terminal UI under /experimental |
GitHub Blog Changelog |
| 2026-06-02 |
Cloud and local sandboxes entered public preview |
GitHub Blog Changelog |
| 2026-06-02 / 2026-06-05 / 2026-06-18 |
GPT-4.1, GPT-5.2/GPT-5.2-Codex, and Opus 4.6 (fast) deprecation notices |
GitHub Blog Changelog |
Hands-on practice
Talks & demos
The three videos below are from the official GitHub YouTube channel and pair directly with this workshop. They are the primary inspiration for the inline "Watch it" callouts in Getting Started, Feature Deep Dive, and Demo Scenarios.
| Resource |
Maps to |
What it covers |
| Ultimate GitHub Copilot CLI tutorial for beginners |
Parts 0–2 |
The official Copilot CLI for beginners series in one video: install & /login, first requests, interactive vs. non-interactive modes, a slash-command tour (/model, /context, /compact, /clear, /resume, /diff, /cwd, /reset-allowed-tools), the plan → delegate → review loop (/plan, /delegate, /review), MCP (/mcp add, Playwright/Svelte), and instructions/agents/skills (/init, .instructions.md, .github/skills, .github/agents) |
| Build with the Copilot CLI — Mona Mayhem |
Part 3 (Demos) |
A GitHub Dev Days hands-on workshop that builds a retro-arcade app end to end: context engineering with /init, plan mode, autopilot vs. YOLO mode, iterative UI debugging by pasting screenshots, the /plugin marketplace + awesome-copilot, /fleet parallel agents, and /delegate to the cloud agent |
| Less // TODO: more done with GitHub Copilot CLI |
Parts 0–2 + SDK |
A Dev Days deep dive into the three interaction modes (non-interactive/headless for CI, interactive, server mode), /login & /user, /env, Auto model selection, built-in agents (research, plan, autopilot), /review, /rubber-duck cross-model validation, /fleet + /tasks, Copilot Memory, and driving the CLI from an app via the Copilot SDK |
| GitHub Copilot Anywhere: From Remote Control CLIs to Cloud Sandboxes (DEM305) |
Sandboxing |
Microsoft Developer demo. Inspiration for the Copilot Anywhere narrative: remote-controlling the CLI (prompt mode, ACP, /delegate) and offloading work to cloud sandboxes you can resume from any device |
Note on workshop source decks
This workshop was informed by internal Microsoft/GitHub enablement decks provided as reference material. Those decks contain forward-looking or templated content (for example, speculative future model names and dated context-window figures) that is deliberately not reproduced here. Every factual claim in this workshop is instead tied to the GitHub primary sources above. Three of the supplied reference files were rights-protected (encrypted) and could not be opened; none of their content is represented here.