Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsAI生成内容标识合规 Gate | China AIGC Labeling
Check AI images, video, audio, text & AI apps against China's AIGC labeling rules (标识办法 + GB 45438-2025) — exact label fixes, metadata check, PASS/REVISE/BLOCK.
$24
AI生成内容标识合规 Gate | China AIGC Labeling
Example session with this skill installed
我们是一个MCN，给品牌做AI数字人带货短视频（竖屏1080×1920，数字人是AI合成的，声音是克隆的主播声音），还有AI生成的海报。视频右下角有一个'AI'小字，大概30像素高，整条视频都有。海报没加标识。我们发抖音和小红书。另外，有些品牌方要求把'AI'字样去掉，说影响转化，能不能导出无标识版本？
[附件：海报 poster.png]
(Fictional test case.)
- Read your context and instructions
- Compiled the ai gate
AI生成内容标识合规检查报告 | MCN 数字人带货视频 + AI 海报
角色： D. 发布者（MCN，代品牌发布AI内容；第十条）。如贵司自研并对外提供数字人生成工具，可能同时属于 A 类，请告知后补充检查。
0. 未执行的请求
| 请求 | 未执行原因 | 合规替代方案 |
|---|---|---|
| 导出去掉"AI"字样的无标识版本 | 《标识办法》第十条：不得恶意删除、篡改、隐匿标识，不得为此提供工具或服务 | 保留标识，优化标识的位置和样式（仍在边角、足够大、清晰）；在合同中写明标识不得删除 |
1. 结论
BLOCK：品牌方要求去标识属于第十条禁止行为；海报既无显式标识也无隐式标识；视频标识缺"生成/合成"要素且字号低于标准。
2. 逐项检查表（节选）
| # | 对象 | 要求（条款） | 发现 | 状态 |
|---|---|---|---|---|
| 1 | 视频 | 含 AI 要素和生成/合成要素（GB 45438-2025 §5） | 仅有"AI" | 不符合 |
| 2 | 视频 | 起始画面、边角、≥ 2 秒（§5.4） | 右下角，全程显示 | 符合（据描述） |
| 3 | 视频 | 文字高度 ≥ 最短边 5%（§5.4） | 1080 × 5% = 54 px；现约 30 px | 低于标准，建议放大 |
| 4 | 视频 | 发布时使用平台AI声明功能（第十条） | 未提及 | 未核实 |
| 5 | 海报 | 显式标识（第四条；§5.2） | 未添加 | 不符合 |
| 6 | 海报 | 隐式标识元数据（第五条；附录E） | find_aigc_metadata.py poster.png：No 'AIGC' marker found | 不符合 |
| 7 | 视频 | 隐式标识元数据 | 未提供文件 | 未核实 |
3. 必改项
| 优先级 | 问题 | 修改方法 |
|---|---|---|
| BLOCK | 品牌方要求去标识 | 拒绝；在合同中加入标识保留条款 |
| P0 | 海报无显式标识 | 在边或角加"AI生成"，文字高度 ≥ 54 px（1080 × 5%），对比清晰 |
| P1 | 视频标识缺要素 | 改为"AI生成"或"本视频由AI生成合成" |
| P1 | 视频标识字号 | 从约 30 px 提到 ≥ 54 px |
| P1 | 隐式标识 | 确认数字人、声音克隆、生图工具导出时是否写入 AIGC 元数据；剪辑转码时不要剥离 |
| P1 | 平台声明 | 发布抖音、小红书时使用平台的AI内容声明功能 |
4. 元数据应有的样子（由生成工具写入）
{"Label":"1","ContentProducer":"【生成服务提供者名称或编码】","ProduceID":"【内容编号】","ReservedCode1":"","ContentPropagator":"【同上或传播方】","PropagateID":"【同上】","ReservedCode2":""}
5. 合同条款草稿（MCN 与品牌方）
"双方确认，本合作中使用人工智能生成合成的内容将依照《人工智能生成合成内容标识办法》及 GB 45438-2025 保留显式与隐式标识，任何一方不得要求删除、隐匿或篡改上述标识。"
6. 限制说明
- GB 45438-2025 的尺寸/时长条款主要约束生成服务提供者；发布者的直接义务是第十条（声明 + 不得去除标识）。
- 克隆主播声音涉及声音权益和授权问题，需另行评估。
- 抖音、小红书可能有更严格的平台规则。本报告为合规预检，不构成法律意见。
Connects securely to your tools. The creator never sees your data.
About this skill
Since 2025-09-01, AI-generated content in China needs labels: a visible one (显式标识) and one in the file's metadata (隐式标识). The rules are specific — the label must say both "AI/人工智能" and "生成/合成", image and video labels must be at least 5% of the shortest side, video labels must show on the opening frame for at least 2 seconds, and the metadata needs an AIGC field with seven keys. In April 2026, press reports said regulators summoned 剪映, 猫箱 and 即梦AI for not implementing them.
自2025年9月1日起，AI生成内容必须标识。这个 skill 对照《人工智能生成合成内容标识办法》和强制性国家标准 GB 45438-2025，逐项检查你的内容或产品，给出具体的改法。
Who it's for | 适用对象
- Creators, brands and MCNs publishing AI content on 抖音, 小红书, 视频号, B站 — declaration duties, keeping labels intact, client contract clauses
- AI app and SaaS builders — explicit labels by media type, metadata, export behavior, user-agreement terms, the "unlabeled on request" flow with 6-month logs
- Platforms hosting user content — the confirmed / possible / suspected notice tiers
- App stores — developer declaration and verification
What you get
- PASS / REVISE / BLOCK verdict, with your role(s) identified first, since obligations differ by role
- Item-by-item check table, every requirement cited to its article or GB clause
- Exact fixes: label wording, placement, and minimum size in pixels with the math shown (1080×1920 → 54 px), video duration
- Metadata check script (stdlib Python, streams large videos) that finds the AIGC field, validates all 7 keys and the Label value, and warns on duplicates
- Drafts: AIGC metadata example, user-agreement labeling clause, platform notice wording, publisher–client contract clause
- Refuses label removal: requests to strip watermarks or export "clean" versions are flagged BLOCK under Art. 10, with a compliant alternative
Every requirement is cited; nothing is added that the rules don't say. Platform-specific rules and portrait/voice rights are flagged for separate review. A compliance pre-check, not legal advice. Current as of October 2026.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
30-day refund
Not a fit? Get your money back
Trust & safety
Security scanned
Verified clean 3 days ago
- Passed all security checks, Safe to install