Skip to content

track-1-copilot-cli.md

Latest commit

 

History

History
100 lines (66 loc) · 5.39 KB

track-1-copilot-cli.md

File metadata and controls

100 lines (66 loc) · 5.39 KB

Track 1 — GitHub Copilot & Copilot CLI

The "can I use it?" track. Foundational tool proficiency for IDE and terminal workflows.

What this track covers

GitHub Copilot in the editor, Copilot CLI in the terminal, MCP (Model Context Protocol) for cross-tool skills, and the discipline of validating AI suggestions with human judgment.

Level 100 — Foundational

Audience: PMs, ops, business managers, anyone new to AI-assisted work.

Hands-on focus: Install GitHub CLI and Copilot CLI, authenticate, use Copilot Chat in VS Code, validate suggestions with human judgment. Practice rejecting bad suggestions as much as accepting good ones.

Outcome: Participant can run Copilot Chat in VS Code and Copilot CLI in the terminal, complete one realistic task (e.g., summarize a document, draft an email, run a CLI command they hadn't memorized), and articulate when they should not trust the suggestion.

Microsoft Learn:

External docs:

Live session structure (60 min):

  • 10 min: What Copilot is (and isn't). The "validate before accept" mindset.
  • 35 min: Hands-on — three real tasks from the audience's daily work.
  • 10 min: Failure-mode discussion — when did Copilot give bad suggestions? What did you notice?
  • 5 min: RAI checkpoint.

Customer-facing notes: For customer workshops, swap "version control and issue trackers" for the customer's actual tooling. K–12 example: lesson planning, parent communication drafts, rubric generation. Healthcare example: patient communication drafts, summarization of public clinical guidelines. The Copilot mechanics are identical — only the scenarios change.

Level 200 — Applied

Audience: PMs, engineers, TPMs integrating Copilot into team workflows.

Hands-on focus: Configure MCP (Model Context Protocol) servers, build cross-tool skills, add explicit approval gates, integrate Copilot with version control and issue trackers.

Outcome: Participant can configure at least one MCP server, build a multi-tool skill that requires explicit approval before action, and integrate Copilot output with a downstream tool (e.g., creating a work item, opening a PR).

Microsoft Learn:

Live session structure (60 min):

  • 10 min: What MCP is — and what cross-tool skills unlock.
  • 35 min: Hands-on — configure one MCP server, build one skill, run it end-to-end.
  • 10 min: Approval-gate design — where in the skill should the human say "yes"?
  • 5 min: RAI checkpoint.

Level 300 — Production

Audience: Engineers, SREs, technical leads deploying Copilot at organizational scale.

Hands-on focus: Build enterprise CLI skill libraries, govern cross-team skills, instrument usage telemetry for ROI reporting, integrate with Microsoft Entra ID for identity-aware skills.

Outcome: Participant can publish a curated skill library to a team, configure identity-aware access (so the skill respects what the user can see), and instrument basic telemetry on skill usage for adoption reporting.

Microsoft Learn:

Live session structure (60 min):

  • 10 min: From "I built a skill" to "the team uses skills" — the governance gap.
  • 35 min: Hands-on — publish a skill, gate access via Entra ID, add telemetry.
  • 10 min: ROI question — what does a "useful" skill look like in your telemetry?
  • 5 min: RAI checkpoint — what does an identity-aware skill see, and what should it not see?

Common pitfalls

  1. Demoing without validating. A facilitator who accepts every Copilot suggestion teaches the wrong lesson. Reject suggestions out loud and explain why.
  2. L100 with no hands-on. A 60-minute Copilot session that doesn't have the audience open Copilot is a webinar, not a workshop.
  3. L300 without identity context. Skills that ignore identity create exactly the data-leak failures Track 0 warned about.

What success looks like

  • 30-day metric: % of participants who used Copilot at least once after the session
  • 90-day metric: workflow change attributable to Copilot use (one specific task, time saved, error reduction)
  • Long-term: organic skill sharing — participants demo their own skills to peers without being asked

Microsoft Learn module titles and URLs occasionally change. If a link 404s, search the module title at learn.microsoft.com and open an issue. The canonical link list with last-verified dates lives in _module-index.md.