claudeemailskill 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 Claude 的邮件工作时使用,涵盖邮件提示、活动起草、评审循环和人工审查。触发场景包括:提示词包、活动草稿评估、评审标准、改写循环以及 Claude Code 邮件交接等请求。
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。
技能指令原文(SKILL.md)
Claude Email Skill
Use Claude for structured reasoning and critique, but keep the message accountable to the brief, audience, and approval owner.
Decision Lens
- Primary lens: Claude-oriented email prompts, campaign drafting, critique loops, and human review.
- Common request signals: prompt packs, campaign draft evaluation, review rubrics, rewrite loops, and Claude Code email handoffs.
- Default posture: Use Claude for structured reasoning and critique, but keep the message accountable to the brief, audience, and approval owner.
- Useful output family: Claude prompts, campaign drafts, critique rubrics, rewrite instructions, variant sets, and approval-ready copy packets.
- Production boundary: separate recommendation from execution.
When To Use
Use this skill for Claude-oriented email prompts, campaign drafting, critique loops, and human review. It is designed for agent workflows where email work must be specific, reviewable, and safe across planning, drafting, implementation, QA, or operational handoff.
Operating Workflow
- Turn the request into a clear email brief with audience, promise, proof, CTA, tone, constraints, and forbidden claims.
- Choose the Claude task shape: ideation, first draft, critique, rewrite, variant generation, or final QA.
- Provide Claude with bounded source material and ask it to cite which input drove each major claim.
- Review outputs against a rubric instead of taste alone: relevance, specificity, risk, readability, and production readiness.
- For rewrites, preserve approved facts and change only the requested dimension such as tone, length, structure, or CTA emphasis.
- Mark anything that needs human approval, legal review, data validation, or ESP testing.
Review Criteria
- The prompt gives Claude enough context without exposing unnecessary private data.
- Drafts do not invent guarantees, customer proof, pricing, or compliance claims.
- Critique feedback is actionable and tied to the brief.
- Variants test a meaningful hypothesis rather than superficial wording.
- Final copy includes subject, preview, body, CTA, and QA notes when relevant.
Output Pattern
Return Claude prompts, campaign drafts, critique rubrics, rewrite instructions, variant sets, and approval-ready copy packets. 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.