Proofread and clean tracking from email drafts

This skill fixes the small mistakes that make an email look careless, while leaving the writer's voice and commitments untouched. Pass 1 corrects objective errors only: duplicated words, wrong articles (choosing a/an by sound, not spelling), subject-verb agreement, wrong-word slips (your/you're, its/it's, than/then, affect/effect, their/there), capitalization, and doubled spaces or punctuation. It explicitly leaves alone anything that is a voice choice rather than an error - sentence fragments, contractions, one-line paragraphs, lowercase greetings and sign-offs, informal phrasing, slang, and repetition used for emphasis - and when it can't tell whether something is a mistake or a deliberate choice, it leaves it and notes it in the summary instead of changing it. It never touches names, numbers, dates, quoted text, or anything that changes what the email promises or asks for.

Pass 2, link cleaning, only runs when the user explicitly asks for it - it is never automatic during a normal proofread. When asked, it strips only a fixed list of unambiguous tracking parameters (utm_source, utm_medium, utm_campaign, utm_term, utm_content, mc_cid, mc_eid, fbclid, gclid, igshid, _hsenc, _hsmi) and keeps every other parameter, since values like id, page, v, or q are often what makes the link function. Generic keys like ref or ref_src are treated as ambiguous and left alone unless context confirms they're pure attribution. It never deletes a whole link or removes a repeated link (signature links and repeated calls-to-action are usually deliberate), and it flags rather than silently fixes shortened links, unfamiliar domains, or a query string that looks like a token, session ID, or credential - handing that decision to the writer. If a parameter does look like a leaked credential or token, the skill redacts the value in its own output (e.g. ?auth=[REDACTED]), never echoes the real value, lists the redaction under "Changes made," and tells the writer to re-enter or rotate the value before sending.

Use it when the user asks to fix, proofread, or check the grammar on an email draft before sending, or separately asks to clean tracking parameters out of its links. Do not use it for casual rewrites, tone makeovers, or structural restyling the user didn't ask for - if the user wants that, the right move is to ask what direction they want first, not to reinterpret a proofread request as a rewrite request.

• Input: an email draft in any format or layout, optionally with a request to also clean tracking links.
• Output: the corrected email in its original format and layout; a "Changes made" list, one line per change; a "Left alone" list for anything judged to be voice or an ambiguous link; and, if nothing needed fixing, a plain statement of that rather than invented changes.

Anyone who wants a pre-send grammar and typo check on an email draft - and, on request, tracking-parameter cleanup on its links - without the tool silently rewriting their voice or altering what the email actually says.