GitHub Copilot CLI & Copilot Chat — Productivity Knowledge Base

A practical, opinionated knowledge base for getting the most out of GitHub Copilot CLI and GitHub Copilot Chat. It focuses on two goals:

1. Autonomous delivery — getting a project or task done end‑to‑end with minimal babysitting.
2. Instruction following — making Copilot reliably do what you actually meant, the first time.

Everything here is distilled from the official GitHub Copilot documentation plus field-tested workflows. Sources are linked in docs/references.md.

---

Start here

If you want to…	Read this
Understand the tools & modes	docs/01-fundamentals.md
Get a task done autonomously	docs/02-autonomy.md
Make Copilot follow your instructions	docs/03-instruction-following.md
Learn the explore → plan → code → verify → commit loop	docs/04-workflows.md
Write great prompts	docs/05-prompting.md
Configure custom instructions, agents, skills, MCP	docs/06-customization.md
Grab copy‑paste commands	docs/07-cheatsheet.md
Avoid common mistakes	docs/08-antipatterns.md
Port a feature from a reference repo into your app	docs/09-port-feature-across-repos.md
Understand loop engineering and how it maps to Copilot CLI	docs/10-loop-engineering.md
See ready-to-use examples	examples/

---

The 60-second summary

To get a task done autonomously:

# 1. Plan first (models succeed more with a concrete plan)
copilot
> /plan Add OAuth2 login with Google + GitHub

# 2. Review/edit the plan, then hand it off
> Accept plan and build on autopilot        # or press Shift+Tab into autopilot

# 3. Or run fully headless / in CI:
copilot --autopilot --yolo --max-autopilot-continues 15 \
  -p "Implement the plan in plan.md, run tests, and fix failures"

To make it follow instructions well:

1. Create .github/copilot-instructions.md with your build/test commands, code style, and workflow.
2. Keep instructions short, concrete, and non-conflicting.
3. Use path‑specific .github/instructions/*.instructions.md for language/folder‑specific rules.
4. State the acceptance criteria and verification command in every non‑trivial prompt.

See the working templates in examples/.

---

Repo layout

copilot-cli-knowledge/
├── README.md
├── docs/
│   ├── 01-fundamentals.md          # tools, modes, keybindings
│   ├── 02-autonomy.md              # autopilot, delegate, fleet, headless/CI
│   ├── 03-instruction-following.md # making Copilot do what you meant
│   ├── 04-workflows.md             # end-to-end recipes
│   ├── 05-prompting.md             # prompt patterns
│   ├── 06-customization.md         # instructions, agents, skills, MCP, hooks
│   ├── 07-cheatsheet.md            # commands & flags quick reference
│   ├── 08-antipatterns.md          # what NOT to do
│   ├── 09-port-feature-across-repos.md  # reference-repo → your-app port recipe
│   ├── 10-loop-engineering.md      # the "loop engineering" term: origin, benefits, patterns
│   └── references.md               # official doc links
└── examples/
    ├── instructions/               # drop-in .github/copilot-instructions.md + path-specific
    ├── agents/                     # custom agent profile example
    └── prompts/                    # reusable prompt files & autonomous task briefs

Contributing

This is a living document. When you discover a workflow that reliably improves autonomy or instruction‑following, add it to the relevant doc with a one‑line rationale. Keep entries concrete and copy‑pasteable.