360 Secure SkillHub
360 安全技能中心

Agentic Video Production Reviewer

该 Skill 以纯指令方式审查 AI 视频制作包,核对目标平台、受众、视频文件、标题、描述、缩略图、字幕和披露要求,检查创意一致性、素材来源、平台元数据与公开措辞,区分质量提示和发布阻断项,并返回 ready、ready_with_notes、blocked 或 do_not_publish 结论;不执行上传、发布、排期或账号操作。

zack-dev-cm
zack-dev-cm
热度 31.61
最近更新 2026-05-16 00:01:48
分类:营销依赖:纯提示词应用场景:社交媒体 / 视频编辑 / 文案创作 / 品牌素材 ...

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add agentic-video-production-publisher
Metadata
name
agentic-video-production-publisher
description
Review AI video production packages for creative consistency, asset provenance, platform metadata, disclosure wording, and publishing readiness. Use when a user asks whether a video package is ready for YouTube, TikTok, or short-form release without uploading or scheduling content.

Agentic Video Production Reviewer

Use this skill to review an AI video production package before a human-controlled publish step. The skill is instruction-only and should produce a readiness verdict, not operate accounts or upload media.

Review Workflow

  1. Confirm the target platform, audience, video file, title, description, thumbnail, captions, and disclosure requirements.
  2. Check that character, setting, music, prompt, source-asset, and generated-asset provenance are documented.
  3. Verify platform metadata for policy-sensitive claims, attribution, affiliate disclosure, copyright risk, and audience settings.
  4. Confirm that private local paths, account names, unpublished prompts, and production notes are not present in client-facing copy.
  5. Separate creative quality notes from release blockers.
  6. Return a release verdict: ready, ready_with_notes, blocked, or do_not_publish.

Boundaries

  • Do not upload, publish, schedule, delete, monetize, or modify any video.
  • Do not operate logged-in YouTube, TikTok, browser, or creator-studio sessions.
  • Do not request credentials, cookies, account recovery steps, billing access, or private analytics.
  • Do not expose internal production notes, local paths, private prompts, or unredacted source-asset links in client-facing output.

Output Shape

Return:

  • Package: files and metadata reviewed.
  • Creative: consistency and quality findings.
  • Policy: disclosure, copyright, platform, and audience risks.
  • Public surface: client-facing wording issues.
  • Verification: checks still needed before manual publish.
  • Verdict: one of ready, ready_with_notes, blocked, or do_not_publish.