Plan a multi-file refactor and review every edit before applying - SkilBrill

GitHub Copilot Training Plan a multi-file refactor and review every edit before applying

GitHub Copilot Training

0% Completed

M1 — The Tool and How It Thinks

What the model sees when it suggests: context window and snippet

Inline completion, chat and the agent modes compared

Where suggestions come from and why they are plausible, not correct

Latency, acceptance rate and what the metrics actually mean

Editor setup across the common IDEs and the command line

The habit shift: reviewing output instead of writing every line

M2 — Inline Completion in Practice

Accepting, rejecting and cycling alternatives without breaking flow

Writing the comment or signature that steers the next suggestion

Boilerplate, migrations and the places it genuinely earns time

When completion interrupts rather than helps

Keeping naming and style consistent with the existing codebase

Measuring real time saved instead of feeling productive

M3 — Chat, Context and Prompting for Code

Structuring a chat request: goal, constraints, target file

Referencing files, symbols and terminals as context

Explaining unfamiliar code without accepting its conclusions

Debugging with chat: hypothesis, evidence, then change

Asking for a design before asking for an implementation

Correcting a wrong answer instead of rephrasing and hoping

M4 — Multi-File Edits and Refactoring

Planning a change that spans several modules

Describing the intended end state rather than each edit

Reviewing a batch of generated edits before applying any

Refactoring patterns: extract, rename, migrate, consolidate

Keeping tests green through a generated change

Reverting a partial application cleanly

M5 — Tests, Documentation and Review

Generating tests that cover behaviour rather than mirror the source

Finding the missing case the generated suite does not cover

Property-based and edge-case thinking for AI-written tests

Documentation that stays accurate after the code changes

Reviewing AI-authored code: the failure modes to look for

Comment threads that explain why, not just what changed

M6 — Security and Enterprise Controls

Common security weaknesses in generated code

Secrets, injection and unsafe defaults to check before commit

Pull request review and AI summaries of a change

Custom instructions, repository guides and commit message style

Policies, content exclusion and what an organisation can restrict

Audit, telemetry and proving usage without reading anyone's code

M7 — Applied Skills

Estimating a task with and without assistance honestly

Untangling a suggestion that was right about the wrong thing

Onboarding to an unfamiliar codebase using chat deliberately

Reducing review load when half the diff was generated

Keeping a team's conventions when everyone accepts suggestions

Deciding what should never be delegated to a model

M8 — Capstone

Take an existing codebase and produce a working change with the tool

Plan a multi-file refactor and review every edit before applying

Generate a test suite, then find and fix the case it missed

Add repository custom instructions and prove they change output

Open a pull request with an AI summary and review it against the checklist

Document review standards and the enterprise policy for the team

Announcements Course Info