Skip to content

08-wrap-up.md

Latest commit

 

History

History
78 lines (58 loc) · 4.44 KB

08-wrap-up.md

File metadata and controls

78 lines (58 loc) · 4.44 KB
Raw

Module 8 — Wrap-up and next steps

← Previous: Managing Copilot's infrastructure

You've worked Copilot CLI through a brownfield, multi-stack codebase, codified your team's conventions, built reusable skills and custom agents, wired up lifecycle hooks, registered LSP and MCP servers, modernized a stack, and packaged the whole setup as a plugin. This module is a quick recap and a pointer to where to go next.

What you will learn

  • A consolidated view of the AI infrastructure you built across the course.
  • How the patterns generalize beyond AssetTrack.
  • Where to go next with Copilot CLI.

Recap

Talking points:

  • Working with Copilot CLI (Module 1) — the harness, agent loop, tool surface, permission model.
  • Building AI infrastructure foundation (Module 2) — exploring a brownfield project, filling doc gaps, generating copilot-instructions.md with /init, scoped .instructions files, the Accessibility Expert custom agent, and the imported make-repo-contribution skill.
  • Enhancing the test suite with remote and delegation (Module 3) — Playwright tests locally, /remote against a hosted environment, /delegate to the Copilot cloud agent, reviewing the resulting PR.
  • Shaping Copilot CLI's lifecycle with hooks (Module 4) — wiring lifecycle hooks so tests, lint, and build feedback flow back to the agent automatically.
  • Adding a new feature (Module 5) — /research for the library choice, /plan + rubber-duck, QA + accessibility custom agents, /fleet for parallel execution.
  • Modernizing apps with Copilot CLI (Module 6) — /lsp across stacks, MCP servers as documentation surfaces, /research for a citation-backed plan, per-stack migrator agents driving the upgrade.
  • Managing Copilot's infrastructure (Module 7) — enterprise custom agents, plugins for distribution, and a custom MCP server exposing the inventory database safely.

Generalizing beyond AssetTrack

Talking points:

  • The same pattern (instructions → custom agents → skills → hooks → LSP / MCP → plugins → research + plan + fleet) applies to any brownfield repo. Walk the learner through how to bootstrap each piece on their own codebase.
  • Common variations:
    • A team conventions copilot-instructions.md shared across many repos.
    • User-level skills for tasks that recur outside any one project (e.g., "rewrite this commit message in our team's house style").
    • Custom agents for compliance-sensitive areas (security review, license audit).
    • Custom MCP servers for the resources your team queries most: databases, internal APIs, feature flags.
  • Knowing when not to use any of this — for genuinely one-off, exploratory work, a plain CLI session is still the right tool.

Suggested next steps

Talking points:

  • Apply the instructions / custom agents / skills / hooks / MCP / plugin pattern to a real repo at work. Start with copilot-instructions.md.
  • Share one custom agent or skill with your team. See how the conversation changes when everyone has the same one available.
  • Explore the MCP registry for a server that fits your team's external tools — or write a small one targeting an internal resource.
  • Subscribe to the GitHub Changelog — Copilot CLI features evolve fast.

Further reading

Resources

← Previous: Managing Copilot's infrastructure