Polish or compose an email draft, passed inline, against the failure modes of AI-drafted mail, and make it read as a person wrote it to someone they like. Standing like-human-do pass on every email.
Resources
3Install
npx skillscat add overseers-desk/skillbooks/labs-edit-email-variants-a1-skills-edit-email Install via the SkillsCat registry.
edit-email
Problem this skill exists to solve
AI-drafted email reads to its sender like an email and to its recipient like a project: paraphrased, hedged, padded, anchored to the drafting session, with asks turned into lists. Worse, a follow-up to someone the sender actually met reads as a pitch to a stranger rather than the next line of a relationship.
Standing step: like-human-do
Every email this skill touches passes through ${CLAUDE_PLUGIN_ROOT}/skills/edit-email/like-human-do.md. This is not optional and not a flag. Read it now and apply it before you finalise.
The general email rules are in ${CLAUDE_PLUGIN_ROOT}/skills/edit-email/email-rulebook.md; apply them as a backstop.
Procedure
- Read
like-human-do.mdandemail-rulebook.md. - Assemble what you know: the recipient, why this email is being sent, and everything in the conversation log that genuinely passed between the sender and this person.
- Write the email directly, applying like-human-do. Do not spawn a subeditor for this variant. Produce the email the sender would actually send to someone they liked.
- Output exactly the line
[[A1]]on its own, then a blank line, then the finished email (header preambleto:/cc:/subject:then the body). Print nothing else. - The skill does not send.