Search your gmail from claude
Claude skill for Gmail is an Agent Skill: a written method your AI applies on demand, in Zealos or in any client that reads SKILL.md.
Claude skill for Gmail is an Agent Skill: a written method your AI applies on demand, in Zealos or in any client that reads SKILL.md.
A Gmail skill tells the model how to search an inbox the way a person would — by thread and by relationship rather than by keyword — which is the difference between a search tool and an answer.
| name | email-rules |
|---|---|
| description | Rules for subject lines, length, structure and calls to action in outbound and reply emails, including when not to send at all. Use whenever drafting or reviewing an email. |
| allowed-tools | none — this skill adds judgement, not reach |
| compatibility | Zealos, Claude Code, Cursor, Codex and any client that reads SKILL.md |
| license | Free to use and edit |
--- name: email-rules description: Rules for subject lines, length, structure and calls to action in outbound and reply emails, including when not to send at all. Use whenever drafting or reviewing an email. --- # Email rules ## Length Under 120 words. Outbound: under 90. If it does not fit, the ask is not clear enough yet. ## Subject lines - 4–7 words, lowercase or sentence case. Not Title Case. - Say the subject, not a tease. "Pricing for 40 seats" beats "Quick question". - Never fake a reply with "Re:". - No emoji, no "[URGENT]", no personalisation tokens visible in the subject. ## Structure 1. **One line of why-you**, specific enough that it could not be sent to anyone else. If it could, delete it — a generic opener is worse than none. 2. **The point**, in one or two sentences. 3. **One ask.** A question they can answer in a sentence, or a specific time. No signature block essay. No attachment on a first email. ## Calls to action Ask for a reply, not a meeting, on a first touch. "Worth a look?" outperforms "Do you have 30 minutes Tuesday?" because the cost of answering is lower. ## When not to send - Nothing has changed since the last email. - You are following up on a follow-up with no new information. - The recipient is on a thread where someone else owes the reply. - The message exists to make the sender feel productive. A skipped email costs nothing. A fourth "just bumping this" costs the account.
What did we promise them in March becomes a question with a quoted answer rather than twenty minutes of scrolling.
Your inbox becomes searchable in the way you actually think about it — by person and commitment, not by remembered phrasing.
Picking up an account mid-renewal, you can reconstruct what was agreed without asking the customer to repeat themselves.
Yes. It is a standard Agent Skills folder, so the same SKILL.md applies wherever it is loaded — the tools differ, the method does not.
Read access to Gmail for anything that searches. A skill narrows an agent's tools and can never widen them, so read-only stays read-only.
Only with the tools the agent already has. Most people start with search and add drafting once they trust the answers.
AI email triage is worth having only when it sorts by consequence rather than by recency: what blocks a deal, what someone is waiting on you for, and what can be read at the end of the week.
/email-rulesA follow-up gets a reply when it opens with the specific commitment made on the call, stays under 120 words, and asks for exactly one thing the recipient can answer in a sentence.
/email-rulesA writing-voice skill is the highest-leverage thing you can give an AI, because it changes every sentence the product produces and it needs no tools, no integrations and no data.
/our-writing-voiceAI writing gives itself away structurally before it does lexically: paragraphs of identical length, lists that always contain three items, and a closing summary that restates a body the reader just finished.
/no-slopAfter-call work is the twenty minutes nobody books: the CRM update, the recap, the follow-up, the task for the thing you promised — and it is the first work to disappear on a busy day.
/call-recap-formatA cold sequence works when each touch carries new information: the second email should not exist unless it says something the first one did not, and most sequences fail this test by touch three.
/email-rules