code-review
Review Orleans changes for layer contracts, invariants, minimal complexity, runtime performance, testing, and focused documentation. Use for GitHub Copilot code review and other review tasks.
ソース情報
- リポジトリ
- dotnet/orleans
- ソースの最終更新活動
- 2026年10月10日 17:29
- 検出された SKILL.md の言語
- 英語
- スター
- 10,904
- フォーク
- 2,141
インストール方法
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
ソースファイルを確認
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
SKILL.md を表示中
SKILL.md
ソースの指示 · 読み取り専用プレビュー- name
- code-review
- description
- Review Orleans changes for layer contracts, invariants, minimal complexity, runtime performance, testing, and focused documentation. Use for GitHub Copilot code review and other review tasks.
# Orleans code review
1. Read root `AGENTS.md` and applicable directory instructions. Apply their
implementation principles alongside standard correctness, security,
compatibility, concurrency, reliability, and testing checks.
2. Establish the change's goal and each affected layer's required and provided
guarantees. Trace callers, ownership, lifetimes, and transitions to verify the
contracts and distinguish proven invariants from assumptions.
3. Check the diff for unnecessary state, abstractions, work, allocations, retained
memory, error handling, and documentation edits. Recommend the simplest complete
correction that preserves required behavior and contracts.
4. Examine behavioral tests and the automated PR coverage comment. Identify
meaningful gaps or regressions in affected code; state when evidence is missing.
Support performance claims with measurements or explicit cost assumptions.
5. Flag additional invariants that could simplify code: "If we can guarantee X,
then we could simplify this by Y." Identify the owning layer, enforcement, and
feasibility; distinguish conditional design suggestions from confirmed
contract violations.
Report actionable findings tied to changed lines, explaining the consequence,
evidence, and correction. Prioritize correctness and meaningful runtime or
maintenance costs over subjective preferences; keep recommendations within scope.
GitHubで見る