Your team gets agentic engineering out of the Copilot licences it already has.

How the training runs in GitHub Copilot

hackers&wizards runs GitHub Copilot training for engineering teams as an in-house team training: two days, remote or on site across Germany, Austria and Switzerland, in German or English, then mentoring while your team applies it to real work. It is our Agentic Engineering for Developers and QA training, run on your Copilot licences.

EUR 1,500 per participant. Team training, EUR 15,000 at most for the two days: up to 12 people attend, and you pay for 10 at most. Net, plus VAT. On site, plus travel at cost.

Teams have run every one of our trainings for engineering, product and leadership in GitHub Copilot.

In 2026 an energy supplier's tech and data teams, about 50 people in every role, trained in Copilot CLI in three remote groups split by stack. An insurance software provider's team of 22 trained two days in GitHub Copilot.

What's different in GitHub Copilot

After the training, your team puts each instruction in a file that the Copilot surface it uses actually reads, and decides per tool what Copilot may do without asking.

• Three kinds of instruction file

  .github/copilot-instructions.md covers the whole repository, .github/instructions/*.instructions.md covers the paths in its applyTo glob, and AGENTS.md, CLAUDE.md or GEMINI.md serve as agent instructions. Copilot CLI and the cloud agent read all three. Copilot Chat in VS Code reads the first two plus AGENTS.md, and Copilot Chat in JetBrains IDEs reads only the first two.

• Custom agents, skills and hooks

  Custom agents are .agent.md files in .github/agents. Skills live in .github/skills, .agents/skills or .claude/skills, so skills built for Claude Code carry straight over. Hooks are JSON files in .github/hooks, used by Copilot CLI and the cloud agent.

• MCP, with GitHub already connected

  Copilot CLI comes with the GitHub MCP server already connected. Add more per person with /mcp add or ~/.copilot/mcp-config.json, or for the whole team with .github/mcp.json in the repository.

• Per-tool approvals, sandbox on request

  Copilot CLI reads and searches freely, and asks before it edits files, runs a command that changes anything or opens a URL. --allow-tool and --deny-tool set per-tool rules for a session, and a deny rule wins even against --allow-all. The local sandbox stays off until you run /sandbox enable or pass --sandbox. --yolo allows every tool, path and URL at once, so it belongs inside the sandbox with bypass requests turned off.

• Your IDE matters

  In our deliveries Copilot's agent ran well in VS Code and poorly in IntelliJ (January 2026). That's why teams on JetBrains IDEs train in Copilot CLI, which reads the same instruction files.

What you set up on your side

• A Copilot seat for every participant. On Business or Enterprise, your organization manages Copilot policies for the whole team. Copilot CLI comes with every Copilot plan, but organization seats need the Copilot CLI policy switched on, so we check it with you before day one.
• Our hosted training environment runs Claude Code, so a Copilot training runs on your machines and licences. Your code stays put: the trainer works with your team on screen.
• Settle with your security team which permissions participants may grant. Some companies rule out --allow-all-paths, and every exercise works without it.

That way seats and permissions are cleared before day one in Copilot starts.

Day one in GitHub Copilot

By the end of the day, your team can check that an instruction takes effect, hand Copilot a spec in plan mode and block pushes to main for good with a hook.

This is how the first day runs in Copilot CLI, taken from a delivered training. Your trainer cuts it to your codebase in the preparation call.

• Watch the instructions take effect

  Participants write a copilot-instructions.md, ask a question, change one rule and ask again, then quit and pick the session back up with copilot --resume. The usual mistake is a rule Copilot never sees because the file is in the wrong folder. The trainer has everyone confirm the answer changed.

• A spec sprint in plan mode

  The team writes the repository instructions, drafts a feature spec with Copilot and hands it over in plan mode, which asks clarifying questions before touching any code. With /model, participants run the same spec through two models and compare.

• Deny first, then let the agent run

  By the end of the day, pushes are blocked and the tools the team approved run without prompts. Participants block every push for the session with --deny-tool='shell(git push)', and block pushes to main for good with a preToolUse hook in .github/hooks, which lives in the repository and applies to every session. Everything else Copilot does under --allow-tool. A review agent in .github/agents checks the result before the pull request.

What stays the same

The practice is the same in every agent: how the agent loop works, what context it needs, how to delegate so it checks its own work, and how the whole team keeps one way of working. The agenda, the trainers and the mentoring are those of the training you book.

Questions before you book

Sources

Vendor facts on this page, checked against the vendor's documentation on 27 October 2026.