GitHub Copilot CLIで絵文字リスト生成ツールを構築する Building an emoji list generator with the GitHub Copilot CLI

GitHub has published a tutorial on its official blog walking developers through the construction of an emoji list generator using the GitHub Copilot CLI, the company's terminal-based AI coding assistant. The piece is positioned as a hands-on demonstration of how engineers can stay inside the shell while delegating code generation, file manipulation, and command execution to an AI agent.

The tutorial covers the full workflow: installing and authenticating the Copilot CLI, crafting natural-language prompts, and verifying the behavior of the scripts the agent produces. The choice of project is deliberately modest. An emoji list generator typically involves pulling Unicode data or querying an emoji-related API, parsing the results, and outputting them in a structured format. That makes it a manageable sample for showing how a CLI agent reasons about external data sources without requiring readers to set up a complex application stack.

According to the post, the workflow proceeds incrementally. The user issues a high-level prompt, and Copilot proposes concrete actions, such as creating files, installing dependencies, or running scripts, before executing them. This step-by-step interaction model is meant to illustrate how the agent decomposes a vague request into discrete shell operations, while keeping the developer in the loop for confirmations. The article highlights how prompts can be refined progressively, with the assistant adjusting its plan as the project takes shape.

The broader context is the rapid emergence of terminal-resident AI coding agents over the past year. Anthropic's Claude Code, OpenAI's Codex CLI, and Google's Gemini CLI have all launched in relatively quick succession, and GitHub Copilot CLI sits squarely within that trend. Compared with editor-integrated Copilot, which lives inside VS Code or JetBrains IDEs, the CLI variant appears better suited to shell-heavy workflows, batch file operations, and integration with CI/CD pipelines. For developers who already orchestrate much of their work from the terminal, an agent that speaks the same language may reduce context switching.

The emoji domain itself has a well-established ecosystem that gives this kind of sample project some real grounding. The Unicode Consortium maintains the canonical emoji data files, including emoji-test.txt and the CLDR annotations, and libraries such as emojibase, node-emoji, and emoji-datasource wrap that data for application use. A generator script of the type built in the tutorial would typically either fetch from these sources or transform their outputs into JSON, Markdown, or HTML. While the project is small, it touches on patterns that recur in many real-world tasks: fetching upstream data, normalizing it, and emitting it in a consumable shape.

That makes the tutorial useful less as a production reference than as an introductory case study in how AI agents handle external data sources. Readers can observe how Copilot decides which library to use, how it structures the output, and how it responds when the initial result does not match the user's intent. The same pattern, scaled up, applies to scraping APIs, generating reports, or building internal tooling.

GitHub Copilot CLI is still relatively new, and some of its capabilities are reportedly in a preview or evolving state. Behavior, command surface, and pricing may continue to shift as the tool matures, and developers experimenting with it should expect occasional rough edges. The competitive landscape also suggests that feature parity across CLI agents, such as background execution, sandboxing, and richer permission models, is likely to be a moving target over the coming months.

For GitHub, the tutorial serves a dual purpose. It onboards developers to a specific product, but it also reinforces a broader narrative the company has been pushing: that Copilot is no longer just an autocomplete inside the editor, but a family of agents that can act across the developer's environment. Whether the CLI version becomes a daily driver for most engineers or remains a complement to editor integrations may depend on how reliably it handles longer, multi-step tasks, an area where all current terminal agents are still being tested in practice.