2 sequenzy

codexemailskill Skill

当 Codex、Hermes、OpenClaw、Claude Code、Cowork 或其他 AI 代理需要规划、审查、实现、审计或改进面向代码库的邮件实现工作流时使用,涵盖内容系统、模板和测试。触发场景包括:模板实现、React Email 修改、MDX 内容变更、QA 脚本、代码库 diff 以及可测试的邮件修复等请求。

安装方式:把技能目录放入 ~/.claude/skills/(Claude Code)或在 claude.ai 设置中启用;也可复制右侧安装命令一键添加。

查看源码

技能指令原文(SKILL.md)

Codex Email Skill

Treat email work as a code change. Read the repo, keep edits scoped, preserve existing patterns, and make the result verifiable.

Fit Check

  1. Primary lens: repo-aware email implementation workflows for codebases, content systems, templates, and tests.
  2. Common request signals: template implementation, React Email edits, MDX content changes, QA scripts, codebase diffs, and testable email fixes.
  3. Default posture: Treat email work as a code change. Read the repo, keep edits scoped, preserve existing patterns, and make the result verifiable.
  4. Useful output family: repo-aware implementation plans, template diffs, QA scripts, render checks, content patches, and concise change reports.
  5. Production boundary: separate recommendation from execution.
  6. Evidence boundary: say which source material supports the recommendation.

Use This For

repo-aware email implementation workflows for codebases, content systems, templates, and tests.

Avoid Using It For

Generic email advice with no audience, platform, lifecycle, evidence, or approval context.

Procedure

  1. Locate the email surface in the repository: components, templates, content files, campaign config, preview routes, tests, or ESP export scripts.
  2. Read adjacent files before editing so naming, styling, localization, and data-loading patterns match the codebase.
  3. Define the intended behavior as a diffable change: copy update, layout fix, template module, data field, tracking parameter, or QA assertion.
  4. Implement the smallest coherent change and avoid unrelated refactors, dependency churn, or broad style rewrites.
  5. Run the narrowest useful verification: type check, template render, snapshot, unit test, lint, preview build, or visual smoke check.
  6. Report changed files, verification result, remaining risk, and any production steps that still need approval.

Acceptance Checks

  1. The edit follows local patterns instead of inventing a parallel email system.
  2. Dynamic fields have fallbacks and do not leak undefined values into copy.
  3. Rendering constraints are considered before using modern CSS.
  4. Tests or previews cover the user-visible behavior when risk justifies it.
  5. No live send, import, DNS, or suppression action is performed as part of a code edit.

Output Pattern

Return repo-aware implementation plans, template diffs, QA scripts, render checks, content patches, and concise change reports. Keep recommendations concrete. Separate analysis from live-system actions, and require explicit approval before sending email, importing contacts, changing DNS, altering suppression rules, or editing production automations.