Proposal: Portable C++ Core Guidelines Agent Skill with rule-linked evaluations

オープン
#2,342 コメント 2 件 リアクション 0 件 担当者 0 名 GitHub で見る

まだ誰も着手していません。

評価

難易度
5/5
見積もり時間
1週間以上
初心者へのやさしさ
35/100
issue の種類
機能追加
明瞭さ
おおむね明確
活発さ
活発
技術スタック
cpp, github
領域
ai, documentation

調査の方向性

Start with the repository’s existing tool support and automatic processing, then review the C++ Core Guidelines source and the Agent Skills specification linked in the issue. The proposal needs a decision on whether the package belongs here or in a companion project, plus scope and maintenance expectations. Done would include an agreed pilot covering 10–20 rules and a plan for rule-linked evaluations, synchronization, and reporting.

索引モデルが issue の本文から書いたものです。

説明

Would the project be open to a portable Agent Skills package derived from the C++ Core Guidelines, accompanied by a vendor-neutral evaluation suite?

The goal would be to help coding agents apply and cite relevant guidelines during implementation and review across compatible development tools. This seems aligned with the project's existing emphasis on tool support and automatic processing.

Proposed approach
  • Keep the existing guidelines as the authoritative source. Generate indexed reference material preserving rule IDs, explanations, exceptions, links, and license attribution, and record the source commit.
  • Provide a concise SKILL.md that guides an agent to inspect the project's C++ version and constraints, load relevant rules on demand, explain findings with rule references, and validate changes using available compiler checks, tests, and static analysis.
  • Document client compatibility and tool prerequisites. Reuse existing checks, including applicable clang-tidy checks.
  • Distinguish reference coverage from evaluated coverage: making the whole document accessible would not imply that every rule can be automatically enforced or has been validated.
Shared evaluations

Publish evaluation cases and expected outcomes independently of any particular model or agent runner. Cases should cover:

  • Genuine violations and correct rule attribution.
  • Compliant code that should not be flagged or changed.
  • Documented exceptions and project-specific constraints.
  • Fixes that compile, pass relevant tests, and preserve intended behavior.
  • Situations where additional context is needed before making a recommendation.

For example, ownership tests should distinguish a raw owning pointer from a legitimate non-owning raw pointer, rather than rewarding blanket replacement with smart pointers.

Compare the same agent with and without the skill, record model/tool versions, and report results per rule, including missed violations, false positives, and incorrect edits. Use held-out cases beyond the examples embedded in the guidelines, with compiler/test checks and human-reviewed expectations where applicable.

Initial scope and maintenance

A first pilot could evaluate 10–20 rules around resource management, ownership, and lifetime. Broader evaluation coverage could grow incrementally based on evidence and review.

Would this be appropriate for this repository, or better maintained as a linked companion project? Feedback on pilot scope, synchronization with upstream changes, and maintenance expectations would help establish a manageable contribution path.

主要言語
CSS
スター
45.5k
フォーク
5.6k
平均マージ
6分
マージ済み PR(30日)
1

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

isocpp/CppCoreGuidelines のほかの issue

isocpp/CppCoreGuidelines の issue をすべて見る

似ている issue

AI Infra & Agents の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。