Set the context
Open a folder, find Copilot Chat, and explain agent mode for mixed experience levels.
GitHub Copilot · 60 minutes
Build the capability.
Use Copilot where the real work happens.
A practical session for people who build dashboards, apps, automation, AI agents, and code—and want to do more meaningful work with GitHub Copilot and VS Code.
Focused on personal and small-team productivity, especially for people who already know their work and want a better workflow, not a software-engineering bootcamp.
The outcome
Bring Copilot a real task.
By the end of the session, you will know how to bring a real work problem into Copilot, use VS Code and agent mode to turn it into a working draft, and validate the result before you trust or share it.
“Capability is built through repetition, context, and review.”
Copilot accelerates work, but the user still owns the judgment, the risk, and the final check.
The 60-minute path
From “I have it installed” to “I can do actual work with it.”
Frame the work
Connect the session to real data cleaning, report automation, and small-team productivity.
Prompt with purpose
Use Goal–Context–Constraints–Check to improve the outcome.
Build in Agent Mode
Reconcile three sources, generate a report, preserve exceptions, and add tests.
Package the capability
Share the script, reusable prompt, runbook, tests, and outputs with the team.
Choose your next step
Leave with a bounded task you can test in the next week.
Five capability moves
Start with the work, not the feature.
Explain unfamiliar data, script logic, or recurring errors.
Create a dashboard, app feature, automation, agent, script, or first useful version.
Clean, normalize, reconcile, or reshape messy data.
Turn a repeated manual sequence into a script or workflow.
Prove the outcome, document decisions, and make it reusable.
Choose a workspace
Use the tool where the work already lives.
{ }
VS Code + Agent Mode
The primary workshop: inspect files, plan, edit, run commands, test, and package the project.
⌕
Copilot Chat in VS Code
Use for explanation, design conversation, and focused work that does not need project autonomy.
◈
Copilot app
Use when broad planning begins outside a project; bring the resulting brief into VS Code.
›_
Copilot CLI
Use when the task is command-first and the terminal is the primary workspace.
A repeatable prompt
Goal · Context · Constraints · Check
A simple pattern for turning a vague task into a useful, reviewable outcome.
Goal
What outcome do I need?
Context
What must Copilot know about the source, audience, or workflow?
Constraints
What must it use, avoid, preserve, or limit?
Check
How will I know the result is correct and safe?
From vague to useful
Give Copilot enough to work with.
Weak
“Build the weekly report.”
Agent Mode must guess the sources, joining rules, risk definitions, output contract, and evidence.Stronger
Goal: Build a repeatable account-health report from CRM, usage, and prior-week exports.
Context: Usage keys may require the approved alias map; a five-person team reviews the result every Friday.
Constraints: Never modify sources or guess an unmatched account; use standard Python and explicit rules.
Check: Preserve exceptions, add tests for risky behavior, run the workflow, and trace report claims to source data.
The live workflow
Multi-source account-health automation.
A five-person team manually combines CRM, product-usage, alias, and prior-week data. Agent Mode turns that work into a repeatable, tested, reviewable capability without hiding unresolved records.
1
Inspect
Read every source and expose reconciliation risks before editing.
2
Build and run
Join approved records, preserve exceptions, and generate actual outputs.
3
Test and review
Prove high-risk rules and look for plausible but incorrect results.
4
Package
Add a runbook and reusable prompt so a teammate can repeat the workflow.
Explore the starter demo workspace →
The first output is a draft
Make it a conversation.
1Ask for a plan.
2Inspect assumptions.
3Request a small first version.
4Test it.
5Share the exact result or error.
6Refine only what needs improvement.
Responsible use
Keep ownership of the result.
Accuracy
Do joins, calculations, and report claims trace back to source records?
Safety
Are source exports immutable and outputs written separately?
Policy
Is the data appropriate for Copilot and the chosen storage location?
Fitness
Do tests, exceptions, and representative inputs cover real conditions?
Confident language is not evidence.
Your next step
Choose a one-week experiment.
Pick a task you already understand, that takes 15–60 minutes, happens more than once, and has a result you can verify.
Keep building
Upcoming GitHub events you can join from anywhere.
Every option below is remote-accessible and free to attend online. Pick one and pair it with the capability you just wrote down.
GitHub Universe 2026
Free virtual pass. Keynotes and sessions on Copilot, agent mode, and agentic workflows, with recordings available on demand afterward.
githubuniverse.com →Microsoft Ignite 2026
Free digital pass. Copilot, agents, and the wider Microsoft tooling the capabilities you build will connect to.
ignite.microsoft.com →Microsoft Reactor — Copilot series
Short, free, hands-on classes on Copilot Chat, agent mode, and MCP servers in VS Code. Live with Q&A, then on demand.
developer.microsoft.com/reactor →GitHub events and webinars
The full calendar of upcoming webinars and recorded sessions, including Copilot bootcamps and certification series.
github.com/resources/events →Dates were accurate at the time of writing. Confirm schedules and registration on each site before the session.
Questions?
Bring a task.
Provide context and boundaries.
Check the result.
Your expertise is what makes Copilot useful.