GitHub Changelog

本日のアップデート

🤖 エージェントセッションへの画像追加
🤖 Copilot コードレビューがエージェントアーキテクチャで動作するようになりました
🤖 Copilot 使用状況メトリクスにユーザーレベルの GitHub Copilot CLI アクティビティが追加
🤖 新しいセッションフィルターでエージェントアクティビティを検出・管理
🤖 GitHub Copilot コーディングエージェントの Jira 連携がパブリックプレビューに
🤖 GPT-5.4 が GitHub Copilot で一般提供開始
🛠️ 階層ビューの改善と Issue フォームでのファイルアップロード
🤖 プルリクエストコメントで @copilot のモデルを選択可能に
🛠️ プルリクエストのマージステータスへのクイックアクセスがパブリックプレビューに

🤖 エージェントセッションへの画像追加

エージェントセッションへの画像追加

概要

GitHub now supports adding images to Copilot agent sessions on github.com, enabling users to kickstart their agent interactions by pasting, dragging, or clicking to attach images directly into the agent prompt. This feature is available wherever agent sessions are initiated, including the recently launched Agents tab in repositories. The change enables multimodal input for the coding agent, allowing users to convey visual context such as UI mockups, error screenshots, architecture diagrams, or design references that would be difficult to describe in text alone.
Users can now add images to Copilot agent sessions on github.com via paste, drag-and-drop, or the image icon.
Image support is available wherever agent sessions are initiated, including the Agents tab in repositories.
The feature enables multimodal input, allowing visual context (mockups, screenshots, diagrams) alongside text instructions.
The 'quick start' capability allows users to initiate agent sessions directly from an image rather than writing detailed text descriptions.
Three input methods are supported: clipboard paste, file drag-and-drop, and click-to-upload via the image icon.

エージェントセッションへの画像追加

影響と注意点

💡 Frontend developers gain a powerful workflow for converting UI mockups and design files directly into code through image-based agent prompts.
💡 Developers debugging visual issues can now share screenshots directly with the agent, reducing the time spent describing visual problems in text.
💡 The coding agent becomes significantly more useful for tasks that involve visual context, expanding its applicability beyond purely text-describable tasks.
⚠️ The article provides no information about supported image formats, maximum file sizes, or resolution limits, leaving practical constraints unclear.
⚠️ Data retention and privacy policies for uploaded images are not specified — users handling sensitive visual content should exercise caution.

エージェントセッションへの画像追加

🤖 Copilot コードレビューがエージェントアーキテクチャで動作するようになりました

Copilot コードレビューがエージェントアーキテクチャで動作するようになりました

概要

GitHub Copilot code review has transitioned from its previous architecture to a fully agentic tool-calling architecture and is now generally available for all Copilot Pro, Pro+, Business, and Enterprise users. The agentic approach enables code review to dynamically gather broader repository context — including relevant code, directory structure, and references — before generating feedback, resulting in higher-quality findings that prioritize correctness and architectural integrity. This GA release follows a public preview announced in October 2025 and represents a fundamental architectural upgrade that leverages GitHub Actions for execution, requiring a one-time self-hosted runner setup for organizations that have opted out of GitHub-hosted runners.
Copilot code review now runs on an agentic tool-calling architecture that dynamically gathers repository context during review, moving beyond fixed-context diff analysis.
The feature is now generally available for all Copilot Pro, Pro+, Business, and Enterprise users, graduating from the October 2025 public preview.
Review quality improvements include higher-quality findings prioritizing correctness and architectural integrity, lower noise, and more actionable guidance.
Agentic code review runs on GitHub Actions, leveraging existing CI/CD infrastructure for execution.
Organizations using GitHub-hosted runners require no setup; those with self-hosted runners only need a one-time configuration.

Copilot コードレビューがエージェントアーキテクチャで動作するようになりました

影響と注意点

💡 All Copilot Pro, Pro+, Business, and Enterprise users immediately benefit from more context-aware, higher-signal code review feedback on their pull requests.
💡 Organizations using self-hosted runners exclusively must complete a one-time setup to continue receiving Copilot code reviews, creating a mandatory action item.
💡 DevOps teams managing GitHub Actions infrastructure should monitor Actions minutes consumption, as agentic code reviews may use more compute than the previous architecture.
⚠️ Organizations using self-hosted runners exclusively must complete a one-time configuration to receive agentic code reviews — this is a mandatory action item, not optional.
⚠️ The article does not specify the impact on GitHub Actions minutes consumption, which could increase costs for organizations on metered plans.

Copilot コードレビューがエージェントアーキテクチャで動作するようになりました

🤖 Copilot 使用状況メトリクスにユーザーレベルの GitHub Copilot CLI アクティビティが追加

Copilot 使用状況メトリクスにユーザーレベルの GitHub Copilot CLI アクティビティが追加

概要

GitHub has expanded Copilot usage metrics to include user-level GitHub Copilot CLI activity data, building on the enterprise-level CLI telemetry released the previous week. This update enables organization and enterprise administrators to view per-user CLI-specific activity including request counts, session counts, and token usage totals through a dedicated API endpoint. The feature addresses a visibility gap in Copilot usage reporting by extending telemetry granularity from aggregate enterprise data down to individual user behavior patterns, enabling targeted enablement, support, and consumption planning.
User-level Copilot CLI activity data is now available through a dedicated API endpoint, expanding the enterprise-level telemetry released the previous week.
Administrators can view per-user CLI request counts to understand interaction frequency.
Session count data is available to measure how many distinct CLI sessions each user initiates.
Token usage totals are exposed at the user level, enabling consumption tracking and cost forecasting.
The data is accessible through GitHub's REST API for Copilot usage metrics (Enterprise Cloud, API version 2022-11-28).

Copilot 使用状況メトリクスにユーザーレベルの GitHub Copilot CLI アクティビティが追加

影響と注意点

💡 Enterprise and organization administrators gain per-user visibility into Copilot CLI adoption, enabling targeted enablement and support strategies.
💡 Finance and procurement teams can use token usage data for more accurate Copilot cost modeling and budget forecasting.
💡 Enablement teams can identify low-adoption users and prioritize training resources to improve CLI onboarding and utilization.
⚠️ The article mentions only an API endpoint — there is no indication that user-level CLI data is available through a web dashboard, limiting accessibility for non-technical administrators.
⚠️ Data retention periods for user-level CLI activity are not specified, leaving uncertainty about historical analysis capabilities.

Copilot 使用状況メトリクスにユーザーレベルの GitHub Copilot CLI アクティビティが追加

🤖 新しいセッションフィルターでエージェントアクティビティを検出・管理

新しいセッションフィルターでエージェントアクティビティを検出・管理

概要

GitHub Enterprise AI Controls and agent control plane has been updated with additional session filters that enable enterprise administrators to discover and manage agent activity more effectively. The new filters allow searching by session status (queued, in progress, completed, failed, idle, timed out, cancelled), repository, and user, supplementing the existing agent and organization filters. This enhancement transforms the agent control plane from a basic visibility tool into a more operational management interface where administrators can quickly isolate specific sessions for investigation, compliance review, or troubleshooting.
Three new session filters added to the Enterprise AI Controls agent control plane: Status, Repository, and User.
The Status filter supports seven lifecycle states: queued, in progress, completed, failed, idle waiting for user, timed out, and cancelled.
The Repository filter enables scoping agent session views to specific repositories within the enterprise.
The User filter allows searching by the user who initiated each agent session.
These new filters supplement existing agent-type and organization-level filtering capabilities.

新しいセッションフィルターでエージェントアクティビティを検出・管理

影響と注意点

💡 Enterprise administrators gain significantly improved operational visibility into agent activity, enabling faster troubleshooting, compliance auditing, and adoption monitoring.
💡 Security teams can now identify unusual agent session patterns by filtering on user and status dimensions, improving threat detection for compromised accounts or unauthorized agent use.
💡 Platform engineering teams can monitor agent reliability by filtering for failed and timed-out sessions, enabling proactive infrastructure improvements.
⚠️ The article does not mention API access for programmatic querying of the control plane data, potentially limiting integration with external monitoring and reporting tools.
⚠️ Session data retention periods are not specified, leaving uncertainty about the availability of historical data for long-term trend analysis.

新しいセッションフィルターでエージェントアクティビティを検出・管理

🤖 GitHub Copilot コーディングエージェントの Jira 連携がパブリックプレビューに

GitHub Copilot コーディングエージェントの Jira 連携がパブリックプレビューに

概要

GitHub has launched a public preview integration that allows Jira Cloud users to assign issues directly to the Copilot coding agent, which autonomously analyzes Jira issue context and generates draft pull requests in connected GitHub repositories. This bridges two of the most widely used developer tools — Jira for project management and GitHub for source control — by letting an AI agent handle the implementation handoff. The feature requires Jira Cloud with Atlassian Rovo enabled, a connected GitHub organization, and the Copilot coding agent already active. It targets repetitive tasks like bug fixes and documentation updates, and is available for GitHub Data Residency customers in any supported region.
Jira issues can now be assigned directly to GitHub Copilot coding agent, triggering autonomous code generation and draft pull request creation.
A new GitHub Copilot for Jira app is available on the Atlassian Marketplace, alongside a companion GitHub App, establishing bidirectional integration.
The agent reads Jira issue descriptions, comments, and context to autonomously implement changes without developer intervention.
Progress updates and clarifying questions from the agent are posted back into Jira via the agent panel, maintaining project management visibility.
Repository-level access controls allow organizations to scope which repos the agent can target from Jira assignments.

GitHub Copilot コーディングエージェントの Jira 連携がパブリックプレビューに

影響と注意点

💡 Development teams using both Jira and GitHub can reduce context-switching overhead by assigning implementation tasks directly from their project management tool.
💡 Engineering managers gain a new delegation pattern: routing well-defined tasks to an AI agent from Jira without requiring developers to manually pick up the issue.
💡 Platform/DevOps teams must account for increased GitHub Actions consumption since the coding agent uses Actions runners for its compute environment.
⚠️ Public preview only — feature behavior, limitations, and availability may change significantly before general availability.
⚠️ Requires Jira Cloud specifically; Jira Server and Jira Data Center are not supported, limiting adoption for on-premises Jira customers.

GitHub Copilot コーディングエージェントの Jira 連携がパブリックプレビューに

🤖 GPT-5.4 が GitHub Copilot で一般提供開始

GPT-5.4 が GitHub Copilot で一般提供開始

概要

OpenAI's GPT-5.4, described as their latest agentic coding model, is now generally available across all GitHub Copilot tiers (Pro, Pro+, Business, and Enterprise) and accessible in every major IDE integration, github.com, GitHub Mobile, GitHub CLI, and the Copilot coding agent. The model demonstrates improved performance in real-world agentic and software development tasks, with enhanced logical reasoning and multi-step, tool-dependent process execution. Enterprise and Business administrators must explicitly enable the GPT-5.4 policy in Copilot settings before users can access it, creating a gated rollout mechanism.
GPT-5.4, OpenAI's latest agentic coding model, is now generally available across all GitHub Copilot plan tiers (Pro, Pro+, Business, Enterprise).
The model is selectable in the model picker across 9+ platforms: VS Code, Visual Studio, JetBrains, Xcode, Eclipse, github.com, GitHub Mobile, GitHub CLI, and the Copilot coding agent.
Enhanced logical reasoning and task execution capabilities specifically optimized for multi-step, tool-dependent agentic processes.
Enterprise and Business plan administrators must enable the GPT-5.4 policy in Copilot settings before users can access the model.
Each IDE has a minimum version requirement: VS Code v1.104.1+, VS 17.14.19+, JetBrains 1.5.66+, Xcode 0.48.0+, Eclipse 0.15.1+.

GPT-5.4 が GitHub Copilot で一般提供開始

影響と注意点

💡 All Copilot users across every plan tier gain access to a more capable model for complex, multi-step coding tasks, directly improving autonomous agent reliability.
💡 Enterprise and Business administrators gain a new governance control point through the GPT-5.4 policy toggle in Copilot settings.
💡 Developers using the Copilot coding agent benefit most, as the model's 'agentic coding' optimization targets exactly this autonomous, tool-orchestrating use case.
⚠️ No quantitative benchmarks or specific performance metrics are disclosed — the claim of 'new rates of success' is qualitative only.
⚠️ Business and Enterprise customers cannot use GPT-5.4 until their administrator explicitly enables the policy, creating a potential adoption bottleneck.

GPT-5.4 が GitHub Copilot で一般提供開始

🛠️ 階層ビューの改善と Issue フォームでのファイルアップロード

階層ビューの改善と Issue フォームでのファイルアップロード

概要

GitHub has shipped three improvements to its Projects and Issues experience: hierarchy view in GitHub Projects now supports sub-issue filtering, deduplication, and persistent expand/collapse state; issue forms gain a new upload field type for requiring file attachments like logs, screenshots, and crash reports upfront; and iteration-grouped project views now support bulk-moving all items from one iteration to another. These changes collectively address long-standing community feedback around project management ergonomics, improving both the planning experience for project managers and the issue submission experience for contributors.
Hierarchy view in GitHub Projects now supports sub-issue filtering using syntax like `sub-issues.is:open` to narrow views to matching sub-issues.
Items appearing in multiple hierarchy positions are now deduplicated in most cases, reducing visual clutter in complex project structures.
Expand/collapse state in hierarchy view is now persisted, maintaining the user's view configuration across page navigations and sessions.
Issue forms support a new 'upload' field type that allows maintainers to require file attachments (logs, screenshots, crash reports) during issue creation.
Bulk iteration moves are available from the group header menu in iteration-grouped views, enabling one-click transfer of all items between sprints.

階層ビューの改善と Issue フォームでのファイルアップロード

影響と注意点

💡 Project managers using hierarchy view gain significantly improved navigation efficiency through filtering, deduplication, and persistent state, reducing time spent re-configuring views.
💡 Repository maintainers can now enforce higher-quality issue submissions by requiring diagnostic files upfront, reducing the back-and-forth typically needed to reproduce bugs.
💡 Scrum masters and sprint planners benefit from the bulk iteration move feature, eliminating the tedious process of individually reassigning items between sprints.
⚠️ Deduplication works 'in most cases,' meaning some edge cases will still show duplicate items — the specific conditions where deduplication fails are not documented.
⚠️ The upload field type's file size limits, accepted file types, and storage quota impact are not specified in the announcement.

階層ビューの改善と Issue フォームでのファイルアップロード

🤖 プルリクエストコメントで @copilot のモデルを選択可能に

プルリクエストコメントで @copilot のモデルを選択可能に

概要

GitHub now allows users to select a specific AI model when mentioning @copilot in pull request comments, giving developers control over which model powers the Copilot coding agent's response. After typing @copilot in a PR comment, a model picker appears in the bottom-left of the comment box, allowing selection from the coding agent's supported models. This feature currently works only on github.com for pull request comments (not review comments), with GitHub Mobile and other comment types planned for future expansion. The change enhances the existing @copilot mention workflow that triggers the autonomous coding agent to make changes in any pull request.
A model picker now appears in the bottom-left of the PR comment box when @copilot is mentioned, enabling per-comment model selection for the coding agent.
Users can select from the Copilot coding agent's supported models before submitting their @copilot request in pull request comments.
The feature works in both Copilot-created and human-created pull requests on github.com.
The coding agent continues to run in its GitHub Actions-powered background environment, with the selected model powering its reasoning.
Currently limited to github.com web interface and pull request comments only — review comments, issues, and GitHub Mobile are not yet supported.

プルリクエストコメントで @copilot のモデルを選択可能に

影響と注意点

💡 Developers gain granular control over which AI model processes their coding agent requests in pull requests, enabling model-to-task optimization.
💡 Teams can standardize on specific models for different types of PR changes (e.g., one model for refactoring, another for bug fixes) based on observed performance.
💡 Organization administrators may need to extend their model governance policies to cover the new PR comment model picker surface.
⚠️ Currently limited to github.com web interface only — GitHub Mobile and CLI are not supported for model selection in PR comments.
⚠️ Only works in pull request comments, not in pull request review comments, creating an inconsistent experience across PR interaction types.

プルリクエストコメントで @copilot のモデルを選択可能に

🛠️ プルリクエストのマージステータスへのクイックアクセスがパブリックプレビューに

プルリクエストのマージステータスへのクイックアクセスがパブリックプレビューに

概要

GitHub is rolling out a public preview that surfaces the pull request merge status indicator at the top of every pull request page, including the new 'Files changed' page. This allows reviewers and PR authors to check merge readiness — blockers, missing approvals, failing checks — from anywhere in the PR experience without scrolling to the merge box at the bottom of the conversation tab. Users still on the classic 'Files changed' page must opt into the new experience to see the merge status from that page. The feature addresses a long-standing workflow friction where developers had to navigate away from their current PR context to check merge readiness.
Pull request merge status is now displayed at the top of every PR page, providing persistent merge readiness visibility without navigating to the merge box.
The merge status indicator is clickable, revealing detailed information about specific blockers including missing approvals, failing checks, and branch protection requirements.
The feature is available on the new 'Files changed' page, enabling code reviewers to check merge readiness without leaving the code review context.
Users on the classic 'Files changed' page must opt into the new experience to see the merge status indicator.
The feature is in public preview status, indicating it is available for testing but may change before GA based on user feedback.

プルリクエストのマージステータスへのクイックアクセスがパブリックプレビューに

影響と注意点

💡 Code reviewers can immediately assess merge readiness when opening any PR page, reducing the workflow interruption of checking the merge box on the Conversation tab.
💡 PR authors gain continuous visibility into what's blocking their PR from merging, enabling faster resolution of blockers during the review cycle.
💡 Teams with complex branch protection rules benefit most, as the persistent indicator surfaces all merge requirements without requiring navigation.
⚠️ Public preview status means the feature's design and behavior may change before general availability.
⚠️ The classic 'Files changed' page does not display the merge status, requiring an explicit opt-in to the new experience.

プルリクエストのマージステータスへのクイックアクセスがパブリックプレビューに

本日のアップデート

🤖 エージェントセッションへの画像追加

概要

影響と注意点

🤖 Copilot コードレビューがエージェントアーキテクチャで動作するようになりました

概要

影響と注意点

🤖 Copilot 使用状況メトリクスにユーザーレベルの GitHub Copilot CLI アクティビティが追加

概要

影響と注意点

🤖 新しいセッションフィルターでエージェントアクティビティを検出・管理

概要

影響と注意点

🤖 GitHub Copilot コーディングエージェントの Jira 連携がパブリックプレビューに

概要

影響と注意点

🤖 GPT-5.4 が GitHub Copilot で一般提供開始

概要

影響と注意点

🛠️ 階層ビューの改善と Issue フォームでのファイルアップロード

概要

影響と注意点

🤖 プルリクエストコメントで @copilot のモデルを選択可能に

概要

影響と注意点

🛠️ プルリクエストのマージステータスへのクイックアクセスがパブリックプレビューに

概要

影響と注意点

Thank You!