Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Course Outline
From Autocomplete to Agent: Grasping the Paradigm Shift
- Understanding the differences between Copilot suggestions and agentic multi-step planning.
- Exploring the architecture of the agent loop: plan, generate, execute, and iterate.
- Reviewing language support and model selection strategies for agent tasks.
- Examining real-world examples: scaling from five-line functions to multi-file features.
Activating Agent Mode in Your IDE
- Steps for activation in VS Code, JetBrains, and Neovim.
- Configuring context windows and selecting model tier preferences.
- Establishing workspace rules and excluding large binary files.
- Distinguishing between Copilot Chat and inline agent workflows.
Multi-Step Planning and Execution
- Prompting Copilot to develop features from start to finish.
- Observing how the agent decomposes tasks across multiple files.
- Reviewing individual steps before applying changes.
- Implementing inline rollbacks when steps deviate from the plan.
Executing Terminal Commands Within the Agent Loop
- Installing dependencies via Copilot's terminal integration.
- Running build commands and interpreting their output.
- Managing environment variables directly from Copilot sessions.
- Understanding safety boundaries: identifying commands that require manual approval.
Test-Driven Development with an Agent
- Generating unit tests from existing source code.
- Guiding test creation using natural language prompts.
- Running test suites and analyzing failure logs within Copilot.
- Refining assertions after encountering edge-case failures.
Navigating Large Codebases Effectively
- Automatically locating cross-file references.
- Refactoring shared utilities using Copilot-guided renaming.
- Simultaneously updating configuration and schema files.
- Preventing context window exhaustion through targeted prompting.
Tailoring Copilot to Team Standards
- Crafting repository-specific instructions in .github/copilot-instructions.md.
- Enforcing naming conventions and architectural patterns.
- Excluding sensitive files and directories from context processing.
- Developing team-specific prompt templates for recurring tasks.
GitHub Copilot Enterprise Governance
- Managing seat allocation, billing, and usage analytics.
- Utilizing audit logs to track Copilot-generated content versus committed code.
- Understanding Microsoft IP indemnity policies and licensing implications.
- Blocking specific file patterns from AI suggestion pipelines.
Debugging with Agent Mode
- Analyzing stack traces collaboratively with the agent.
- Employing hypothesis-driven debugging: asking Copilot for failure reasons.
- Using agent-assisted bisect techniques to identify regression sources.
- Mitigating hallucination risks when debugging unfamiliar code.
Performance Optimization and Limit Management
- Comprehending daily request limits and model quotas.
- Optimizing prompt length to prevent truncated responses.
- Switching between models based on specific task requirements.
- Monitoring agent latency and leveraging caching strategies.
Security and Compliance for Enterprises
- Clarifying data handling: distinguishing between data leaving the repository and local storage.
- Preventing the leakage of secrets and credentials through prompts.
- Ensuring compliance with GDPR, SOC 2, and FedRAMP requirements.
- Red-teaming generated code for injection vulnerabilities.
Troubleshooting Common Scenarios
- Investigating why Copilot might ignore codebase context.
- Resolving indexing failures in large repositories.
- Addressing rate limit errors during peak usage hours.
- Fixing IDE extension synchronization issues.
Summary and Future Roadmap
- Recap of Agent Mode capabilities and practical workflows.
- Overview of GitHub's Copilot roadmap and upcoming agent features.
- Resources for staying updated with the latest Copilot releases.
Requirements
- Prior experience in object-oriented or functional programming.
- An active GitHub account and foundational knowledge of Git workflows.
- Familiarity with at least one Integrated Development Environment (IDE), such as VS Code, JetBrains, or Neovim.
Target Audience
- Developers currently using Copilot who aim to unlock its full agent capabilities.
- Engineering managers overseeing the deployment of Copilot across development teams.
- Security teams responsible for reviewing policies on AI-assisted code generation.
21 Hours
