GitHub公布Copilot CLI指令速查表

• Slash commands are prefixed with / and instantly trigger context-aware actions within the CLI environment
• Core shortcuts include Esc (cancel operation), Ctrl+C (cancel/clear/exit), Ctrl+L (clear screen), @ (mention files), / (show commands), and arrow keys (navigate history)[5]
• Non-interactive mode supports programmatic usage via -p flag for single prompts and -s flag to output only Copilot's response, enabling CI/CD pipeline integration and scripted automation[5]
• Context window management uses automatic compression when approaching 95% token limit and warns when less than 20% remains[7]
• Skills can be manually invoked via slash commands (e.g., /Markdown-Checker check README.md) or automatically triggered when relevant to current task[8]
• Hooks provide programmable control at session lifecycle points (preToolUse, postToolUse, userPromptSubmitted, sessionStart, sessionEnd, errorOccurred, agentStop, subagentStop) for enforcing guardrails, logging, and policy checks[8]
• Custom slash commands can be created via markdown files in designated folders, allowing teams to extend functionality beyond default system prompts[6]

The formalization of slash commands in GitHub Copilot CLI represents a shift toward explicit, auditable AI-assisted development workflows. By prioritizing keyboard-driven, repeatable commands over natural-language prompts, GitHub is addressing enterprise concerns around compliance, security, and predictability in AI-assisted coding. The extensibility of custom commands and hooks suggests a future where development teams can enforce organizational policies and guardrails directly within terminal-based AI workflows, potentially reducing friction in adopting AI tools in regulated industries. Integration across Copilot Chat, CLI, and agent mode indicates GitHub's strategy to create a unified AI assistant ecosystem that maintains consistency across different developer interfaces.