Skip to main content

verification

Use before telling the user a Creatify job is done. Covers how to confirm a long-running generation actually finished with a usable result, using the matching *_get poll tool, and how to report failures honestly instead of claiming false success.

来源信息

仓库
creatify-ai/Creatify-agent-plugin
最近来源活动
2026年7月16日 07:20
检测到的 SKILL.md 语言
英语
星标
1
分支
0

安装方式

默认使用会先检查来源的 Prompt;你也可以切换为直接命令,或下载本地副本。

检查来源文件

决定是否安装前,请先阅读 SKILL.md,以及 SkillsMP 当前展示的配套文件。

正在显示 SKILL.md

SKILL.md
来源说明 · 只读预览
name
verification
description
Use before telling the user a Creatify job is done. Covers how to confirm a long-running generation actually finished with a usable result, using the matching *_get poll tool, and how to report failures honestly instead of claiming false success.
# Verification Creatify generation is asynchronous. A create/generate call returning an id means the job **started**, not that it succeeded. Never report success from the create call alone. ## Confirm before claiming done 1. Poll the matching status tool until it reports a terminal state: - Aurora video → `aurora_get` - Custom avatar → `avatar_create_get` - Voice clone → `voice_clone_get` - Asset (image/video) → `asset_get` - Creative agent → `creative_get_session` 2. Confirm the terminal state is **done/success** and that it returns a usable output (a video/image/audio URL, a ready `avatar_id`, a session output). 3. Confirm the result matches what was asked (right subject, right script, right media type/format). ## Report honestly - If the job failed, report that it failed and the reason — do not present a partial or fabricated result as success. - If it's still running after reasonable polling, say it's still processing and keep polling or hand back a way to check; don't guess the outcome. - Return the actual output URL so the result is usable, even when a host also renders an inline preview. Do only what was asked: verify the specific requested output, don't silently add extra generations to "prove" it.
在 GitHub 查看