---
name: check-before-delivery
description: "Use when an AI beginner is about to send, submit, or claim completion of a real document, file, image, slide deck, page, or other deliverable. Do not use for a trivial chat answer with no deliverable. Inspect the actual result, report the most important problems and unknowns, and avoid false completion."
---

# 交付前检查 / Check Before Delivery

在把结果交出去或说“已经完成”之前，打开真实产物做一次轻量检查。文件存在不等于能用。

Before sending a result or saying it is done, open the real deliverable and run a lightweight check. File existence alone is not enough.

## 怎么做 / What to do

1. 找到使用者真正会收到的文件、页面或内容，并尝试打开或查看。  
   Locate the exact deliverable the recipient will receive and try to open or view it.
2. 按对象只检查相关项：  
   Check only what matters for the artifact:
   - 文字：事实、必需信息、可读性。 / Text: facts, required information, readability.
   - 视觉：是否正常显示、能否读清、有没有裁切或明显错位。 / Visual: rendering, readability, clipping, obvious layout errors.
   - 文件：格式、版本、可打开性、文件名和必要附件。 / File: format, version, openability, filename, required attachments.
   - 链接或路径：接收者是否真的能访问。 / Links or paths: whether the recipient can actually access them.
3. 对照这次任务最重要的目标，不套用与当前产物无关的检查项。  
   Compare against the task's main goal and skip irrelevant checks.
4. 不能亲自验证的内容明确写“未检查”，不要猜成通过。  
   Mark anything you could not verify as unchecked instead of guessing.

## 默认输出 / Default output

先给针对本次日常用途的结论：**可以交付 / 暂不建议交付 / 部分可用**。然后只列：

1. 最重要的1—3个问题；
2. 仍然未知的内容；
3. 一个最小下一步。

Start with **ready / not ready / partly usable for this everyday use**, then list only the top one to three problems, remaining unknowns, and one next step.