VS Code の GitHub Copilot で「コミットメッセージ生成」が失敗する場合の対処法A known bug in VS Code's GitHub Copilot causes commit message generation to…

If you have configured GitHub Copilot in Visual Studio Code to write commit messages that match your team's conventions, only to find that the "Generate Commit Message" button suddenly stops returning anything, you are not misconfiguring the tool. According to reports circulating among Copilot users, this silent failure is a known bug that appears when custom instructions for commit message generation are added to settings.json. Because commit message automation is one of the more widely used everyday features of Copilot inside the Source Control panel, a failure that produces no error and no output can be particularly confusing for developers who assume they have made a syntax mistake.

The scenario is straightforward. A developer wants commit messages to follow a shared standard, such as Conventional Commits or an internal style guide, and so they add a custom instruction to settings.json. The relevant setting is the one governing commit message generation, referenced in the source as github.copilot.chat.commitMessageGeneration. Once that individual setting is populated, pressing the generation button no longer produces a message. Crucially, the operation does not throw a visible error; it simply returns nothing, which is why the behavior is described as a silent failure rather than an outright crash. The article that surfaced this issue emphasizes that the root cause lies in this specific configuration key rather than in the wording of the instruction itself, meaning that even a correctly written instruction can trigger the problem.

This distinction matters because the natural response to a non-working feature is to assume the input is wrong. Developers may rewrite their instruction, escape characters differently, or reload the window repeatedly, none of which addresses a defect that sits in how the setting is processed. Understanding that the problem is a known bug reframes the fix as a workaround rather than a correction, and the source material indicates that the accompanying write-up walks through both the cause and a practical way around it.

Some background helps explain why this feature exists and why the failure is disruptive. GitHub Copilot integrates with VS Code's Source Control view, where a small sparkle icon lets users generate a draft commit message based on the currently staged changes. The model reads the diff and proposes a summary, which developers can accept, edit, or discard. Custom instructions are a broader Copilot mechanism that lets teams shape the assistant's output, and VS Code supports several avenues for supplying them, including settings.json entries and, more recently, instruction files placed in the repository. The intent is to make generated content, whether chat responses, code, or commit messages, adhere to organizational norms without repeating the same guidance every time.

Team-level commit conventions are common in professional workflows precisely because consistent messages improve the readability of a project's history, support automated changelog generation, and make tools like semantic versioning easier to apply. That is why so many teams reach for the commitMessageGeneration customization in the first place. When the very setting meant to enforce those conventions disables the feature, the impact falls on exactly the users trying to follow best practices.

For those affected, the general shape of a workaround is to avoid populating the problematic setting in the way that triggers the failure while still conveying the desired conventions to Copilot. VS Code's instruction file support and other configuration surfaces provide alternative locations for the same guidance, though the linked article is the authoritative source for the exact steps it recommends. As with any bug tied to a specific version, the behavior is likely to shift as VS Code and the Copilot extension are updated, so checking the current release notes and the relevant GitHub issue trackers is advisable before assuming the problem persists.

More broadly, the episode is a reminder that AI-assisted developer tools remain fast-moving software, and that customization surfaces such as settings.json can interact with features in ways that are not always well documented. Silent failures are especially costly because they offer no diagnostic signal, and users are left to distinguish between their own errors and defects in the tooling. Consulting community write-ups, and confirming whether a reported issue is officially acknowledged, remains a practical habit when a Copilot feature behaves unexpectedly after a configuration change.