感謝您提交詢問!我們的一位團隊成員將在短時間內與您聯繫。
感謝您提交預訂!我們的一位團隊成員將在短時間內與您聯繫。
課程簡介
From Autocomplete to Agent: Understanding the Shift
- Differences between Copilot suggestions and agentic multi-step planning.
- Architecture of the agent loop: plan, generate, execute, iterate.
- Language support and model selection for agent tasks.
- Real-world examples: from five-line functions to multi-file features.
Enabling Agent Mode in Your IDE
- Activation in VS Code, JetBrains, and Neovim.
- Configuring context window and model tier preferences.
- Setting workspace rules and ignoring large binary files.
- Managing Copilot Chat versus inline agent workflows.
Multi-Step Planning and Execution
- Prompting Copilot to build a feature end-to-end.
- Observing the agent break tasks into steps across files.
- Reviewing each step before applying changes.
- Using inline rollback when steps drift off course.
Terminal Commands Inside the Agent Loop
- Installing dependencies through Copilot terminal integration.
- Running build commands and interpreting output.
- Managing environment variables from within Copilot sessions.
- Safety boundaries: what commands require manual approval.
Test-Driven Development with an Agent
- Generating unit tests from existing source code.
- Driving test creation with natural language prompts.
- Running test suites and interpreting failure logs inside Copilot.
- Refining assertions after seeing edge-case failures.
Navigating Large Codebases
- Finding cross-file references automatically.
- Refactoring shared utilities with Copilot-guided renames.
- Updating configuration files and schema files in tandem.
- Avoiding context window exhaustion with targeted prompts.
Customizing Copilot for Team Standards
- Writing repository-specific instructions in .github/copilot-instructions.md.
- Enforcing naming conventions and architecture patterns.
- Excluding sensitive files and directories from context.
- Creating team-specific prompt templates for common tasks.
GitHub Copilot Enterprise Governance
- Seat allocation, billing, and usage dashboards.
- Audit logs: tracking what Copilot generated versus what was committed.
- Microsoft IP indemnity policies and licensing implications.
- Blocking specific file patterns from AI suggestion pipelines.
Debugging with Agent Mode
- Reading stack traces together with the agent.
- Hypothesis-driven debugging: asking Copilot why a test failed.
- Using agent-assisted bisect to find regression sources.
- Managing hallucination risks when debugging unfamiliar code.
Performance and Limit Management
- Understanding daily request limits and model quotas.
- Optimizing prompt length to avoid truncated responses.
- Switching between models for different tasks.
- Monitoring agent latency and caching strategies.
Security and Compliance for Enterprises
- Data handling: what leaves your repository and what stays local.
- Preventing leakage of secrets and credentials via prompts.
- Compliance with GDPR, SOC 2, and FedRAMP requirements.
- Red-teaming generated code for injection vulnerabilities.
Troubleshooting Common Scenarios
- Why Copilot sometimes ignores your codebase context.
- Resolving indexing failures for large repositories.
- Handling rate limit errors during peak hours.
- Fixing IDE extension sync issues.
Summary and Future Roadmap
- Recap of Agent Mode capabilities and practical workflows.
- GitHub's Copilot roadmap and upcoming agent features.
- Resources for staying current with Copilot releases.
最低要求
- Experience with object-oriented or functional programming.
- A GitHub account and foundational knowledge of Git workflows.
- Familiarity with at least one Integrated Development Environment (IDE) such as VS Code, JetBrains, or Neovim.
Audience
- Developers currently using Copilot who wish to unlock Agent Mode capabilities.
- Engineering managers overseeing the rollout of Copilot across development teams.
- Security teams evaluating policies for AI-assisted code generation.
21 小時