Agent skill

claude-code-coagent

Use claude_code MCP as a third-party co-agent to gather alternative opinions, unblock stalemates, or delegate scoped tasks (reviews, ideas, draft edits) while you remain the decision-maker. Trigger when seeking another perspective or backup executor: "他の意見", "別の人", "違う観点", "打開策ない?", "別LLMの見解", "claudeに頼んで".

View SKILL.md on GitHub Repository
Stars 0
Forks 0

Install this agent skill to your Project

npx add-skill https://github.com/yurake/skills/tree/main/product/claude-code-coagent

SKILL.md

Claude Code Co-Agent

What this skill is for

  • Pull in claude_code as an external reviewer/ideator/executor while you (Codex) keep ownership.
  • Typical asks: alternative solutions, risk review, unblock ideas, scoped edits/tests in bounded paths.
  • Use when the main pair (user + Codex) is stuck or wants a second opinion or fast trial edit.

Roles

  • You (Codex): own context, guardrails, final decisions, and integration of any changes.
  • claude_code: provides opinions, options, or scoped file edits; must work within explicit bounds.

Workflow

  1. Frame the ask
    • Goal and expected output format (e.g., 3 options ranked, risk bullets, patch for file X only).
    • Scope: allowed files/dirs, banned actions, time/complexity constraints.
    • Context: minimal code snippets/diffs needed to answer; avoid oversharing.
  2. Invoke claude_code with a structured prompt (see templates below).
    • タイムアウトしたら1回リトライ。成功しない場合はユーザーへ報告。
  3. Receive and digest
    • Summarize response in 2–3 bullets。
    • MUST: 返答内容を自分の結論・回答に反映する。統合しない場合は理由を明記。
    • Decide: adopt / partial / discard.
    • If edits returned, review diff → apply → run relevant checks/tests if riskful.
  4. Follow-up if needed
    • Ask to narrow to one option, or request a patch limited to specific paths.
  5. Record briefly
    • What was asked, key takeaways, what was applied (optional but recommended for continuity).

Prompt templates

  • Opinion/ideas: "目的: . 制約: . 求める出力: 選択肢3件を重要度順で。禁止: ."
  • Review: "次のdiffを重大リスクがある点のみ箇条書きで。無害な指摘は不要。<diff/snippet>"
  • Scoped patch: "ファイル限定: . 仕様: . 期待: 完全なパッチのみ。禁止: 依存追加/構造変更。"
  • Unblocker: "今の案が行き詰まり。打開策3件、各1行で。前提: ."

Safety/guardrails

  • Always state allowed paths and banned actions before asking for edits.
  • Prefer getting reasoning or options first; request patches only after scoping.
  • Verify diffs yourself; run focused tests when behavior can change.
  • タイムアウト時は1回リトライし、成功した返答は必ず要約して結論に反映。統合しない場合は理由を明示。リトライ失敗時はユーザーへ報告。

References

  • Prompt examples: references/prompts.md
  • Call checklist: references/checklist.md

Recommended Agent Skills

Expand your agent's capabilities with these related and highly-rated skills.

yurake/skills

issue-create

GitHub Issue を作成する手順。依頼を受けたとき、.github/ISSUE_TEMPLATE の内容に応じてテンプレを選び、必須項目を埋めて GitHub MCP（issue_write/create）で Issue を作成する。

0 0
Explore
yurake/skills

ci-fail-followup

PR 作成後に .github/workflows/ci.yml が失敗したときの対応手順。ローカルで pytest 全ケースを実行し、エラー有無で分岐。エラーがあれば修正して commit-push スキルでコミット・プッシュ、無ければ GitHub Actions のログを確認して原因を特定する。

0 0
Explore
yurake/skills

review-rules

セッションの会話から恒久的な振る舞いルールを抽出し、AGENTS.md と突き合わせて追加すべきものを提示する。出力は rules と agents_md_patch の二部構成。

0 0
Explore
yurake/skills

pr-create

PR 作成手順を自動化するためのガイド。依頼を受けたとき、または 参照しているTODOファイル の「PR 作成」以外が完了したタイミングでタイトル・本文整形、テンプレ適用、MCP での PR 作成を行う。

0 0
Explore
yurake/skills

architecture-patterns

Implement proven backend architecture patterns including Clean Architecture, Hexagonal Architecture, and Domain-Driven Design. Use when architecting complex backend systems or refactoring existing applications for better maintainability.

0 0
Explore
yurake/skills

test-risk-analyzer

あらゆる言語/スタックで潜在バグリスクを洗い出し、追加すべきテストケースを提案する。未テストのエッジケース探索や安全性確認、カバレッジ強化のテスト案提示を求められたときに使用する。

0 0
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results