Skip to content
GitHub Copilot Train-the-Trainer
Slides

Work effectively in the GitHub Copilot app

Session M7-01 | Product and Delivery Teams | Beginner


Work in the GitHub Copilot app

Work in the GitHub Copilot app with repository context. Save work that others can review.

Delivery Decision Studio is the six-session path. Service Request Portal is the initiative.


Save work beyond the chat

Use the chat to work through an idea. Keep project context, source references, issues, and saved work available for review after the conversation ends.

Gather evidence → form a proposal → review it → save the decision → verify it

Check Copilot's explanation against the source facts.


Context has a cost

Context choice When it helps
One named file The answer depends on a specific rule or API
A project or repository The task needs relationships across files
An issue or pull request The task needs delivery history and discussion
No repository context The work is general and contains no project facts

Tell Copilot why each source matters. Do not make it guess.


Separate facts, assumptions, and decisions

Type Example
Fact "Issue 42 names a response-time problem."
Assumption "The mobile client uses the documented endpoint."
Decision "The product owner will confirm the client behavior."

Separating these categories makes uncertainty visible.


Know the main work areas

Area Use it for
Chats Exploration and interviews
Projects Repository-grounded work
My work Issues, pull requests, and sessions
Search Finding repository information
Automations Recurring or on-demand tasks
Customize Skills, agents, MCP servers, and canvases

Context comes before the prompt

Use only permitted sources. Attach the smallest set that answers the question.

Name the files, issues, or repository that should support the answer.


Choose the session mode

Mode Use in this path
Interactive Questions, interviews, and reviewed drafts
Plan Review a multi-step approach before changes
Autopilot Work with agreed scope and permission to act without review of every action

Session M7-01 uses Interactive, then Plan. The learner reviews every write.


Use context controls

  • @ adds a file.
  • # references an issue.
  • / opens commands.
  • /context shows current context usage.
  • /compact summarizes a long active session.
  • /clear or /reset starts fresh.

Ask for sources

Explain the initiative for a product owner.
Cite the supplied source for each factual claim.
Mark missing information as unknown.

A confident answer without a source is still unverified.


Inspect what Copilot can use

  • Skills provide repeatable procedures.
  • Agents carry role-specific instructions and tools.
  • MCP servers connect Copilot to systems.
  • Plugins install reusable packages that can include MCP servers and other customizations.
  • Canvases provide visible shared state.
  • Automations repeat tasks with a defined scope.

Availability depends on the plan, policy, and installation.


Use one repeatable work pattern

Read -> propose -> review -> write -> verify

Every later Module 7 session uses this pattern.


Save the working brief

The lab creates initiative-context-brief.md with:

  • facts and sources;
  • unknowns;
  • useful GitHub artifacts;
  • the next decision and owner.

The brief must make sense without the chat history.


Revise the brief when sources change

Initial source
  → first brief
  → approved late evidence
  → focused revision
  → fresh-session verification

Keep the brief consistent with source information the owner has approved.


Lab

  1. Open the repository as a Copilot project.
  2. Inspect the available capabilities.
  3. Start fresh without the source, then compare an unscoped answer with a grounded answer.
  4. Create and commit the initiative context brief.
  5. Apply approved late evidence.
  6. Review and commit the focused revision.
  7. Ask /rubber-duck for independent critique.
  8. Check the brief in a fresh session.

Stop if the app or repository context is unavailable.