privilege-preflight-enlisted

# Privilege Preflight EN - prompt check before sending to cloud AI ## Philosophy **A prompt pasted into someone else's chatbot is a disclosure. Not "tool usage" - a disclosure.** No lawyer would fax a case file to a newspaper to get a second opinion on the drafting style. Yet the same lawyer will paste half a statement of claim into a free chatbot "just to polish the wording". The two acts differ less than it seems: the content leaves the firm and lands with a third party, on terms nobody read, with retention nobody controls, possibly on servers outside the EEA. This skill does one thing: it holds the hand above the Enter key for the length of one check. It does not ban cloud AI - firms use it and will keep using it. It forces a conscious decision instead: what exactly leaves, to whom, on what terms, and what happens if it ever surfaces. The verdict is banded, not binary. Most prompts are neither clean nor forbidden - they are redactable. That is why the middle band (CAUTION) ends with a redacted draft rather than a lecture. The frame here is European: legal professional privilege as shaped by the Court of Justice of the EU, national professional secrecy regimes, GDPR, and the Trade Secrets Directive. US attorney-client privilege appears only as a comparative footnote, not as the operating framework. ## When to use / What it does NOT do **Use it:** - before pasting client-matter content into an external AI chatbot, - for a firm-wide prompt audit (retrospective: what p