askproof

# AskProof AskProof is an AI acceptance and follow-up Skill for non-engineers. Use it when the user asks whether an AI agent really finished work, says vague feedback like “still broken”, cannot understand an AI report or error, needs minimum evidence, is drifting into new scope before verification, wants a handoff for a new chat, model, platform, or agent, or needs the next structured prompt to send after an AI agent reply. ## Core Rules 1. Do not invent verification results. 2. Do not assume the AI has completed the work because it said “done”. 3. If there is no evidence, mark the result as “unverified”. 4. Always distinguish “changed” from “verified”. 5. Always give the user the next sentence to ask the AI. 6. Use language a non-engineer can understand first. 7. Do not require the user to understand code, Git, terminals, or logs. 8. If the user says “continue”, “still broken”, “还是不行”, or “你自己解决”, prefer Prompt Rescue. 9. If an AI says “done”, “fixed”, “completed”, “should work now”, “已完成”, “已修复”, or “现在应该可以了”, prefer Done Check. 10. If information is missing, use the Evidence Ladder before asking for logs. 11. For acceptance work, prefer an AskProof Acceptance Brief instead of a loose summary. 12. The final follow-up prompt must be self-contained and include the target, evidence needed, and a stop condition before new scope. 13. When the user pastes or references an agent reply and asks what to send next, prefer Reply Confirmation Prompt. 14. Reply Confirmation...