Fake follower audit
Skill creativault/skills/creator-scraper-cv/audit/fake-follower-audit
通用AI助手，为creativault打造的技能包
npx -y skills add creativault/skills --skill fake-follower-auditAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
CreatiVault official fake-follower and engagement-quality audit skill. MUST be used when the user wants to inspect one TikTok, Instagram, or YouTube creator for fake followers, suspicious engagement, bot/template comments, audience authenticity, interaction quality, creator risk, or account health. Accepts a creator profile URL or platform plus username/user ID and calls CreatiVault OpenAPI through scripts/fake_follower_audit.mjs. Do not infer fake-follower risk from public web pages or follower counts alone. Use when: fake follower audit, fake followers, follower authenticity, engagement authenticity, suspicious followers, bot comments, creator risk, account health, 假粉检测, 假粉率, 粉丝真实性, 互动真实性, 刷粉, 刷量, 机器评论, 达人风险, 账号健康度, 检测这个达人有没有假粉.
SKILL.md
4.3 KB, as published. Nobody here has run it
Fake Follower Audit（达人假粉检测）
调用方式
对单个达人执行同步检测：
# 方式一：达人主页链接
node ../../scripts/fake_follower_audit.mjs '{"profile_url":"https://www.tiktok.com/@creator","lang":"cn"}'
# 方式二：平台 + 达人标识
node ../../scripts/fake_follower_audit.mjs '{"platform":"instagram","platform_user_id":"creator","lang":"en"}'
profile_url 与 platform + platform_user_id 必须且只能选择一组。
参数
| 参数 | 类型 | 必填 | 说明 |
|---|---|---|---|
profile_url | string | 条件必填 | TikTok / Instagram / YouTube 达人主页链接 |
platform | string | 条件必填 | tiktok / instagram / youtube |
platform_user_id | string | 条件必填 | username / union_user_id / sec_user_id |
service_level | string | 否 | S1 / S2 / S3，后端默认 S1 |
lang | string | 否 | cn / en；Navos 国内默认 cn，海外和 common 默认 en |
不要把视频 URL 当作 profile_url。用户只提供视频链接时，先从链接或已有数据中确认达人主页/标识；无法确认时向用户澄清。
结果字段
| 字段 | 说明 |
|---|---|
quality_score | 互动质量分，0～100，越高越健康 |
fake_follower_rate | 假粉率，小数形式；0.12 表示 12% |
risk_level | low / medium / high / critical |
conclusion | 检测结论，语言由 lang 控制 |
abnormal_types | 异常类型列表 |
signals | 粉丝、互动率、播放粉丝比和评论质量等检测信号 |
creator_profile | 昵称、头像、粉丝数、关注数及受众画像 |
partial_result | 是否因评论、内容或达人资料不足而降级 |
warnings | 降级和数据缺失告警 |
强制解释规则
fake_follower_rate是模型和规则综合估算值，不是平台官方粉丝逐账号审计结果。- 百分比展示必须乘以 100，例如
0.2395展示为23.95%。 quality_score越高越健康，不能反向解释。risk_level=critical也可能表示资料不足导致无法有效检测；必须同时阅读conclusion、partial_result和warnings。partial_result=true时明确标注“部分数据可用”，不要把结果描述为完整审计。- 不要根据粉丝量、单条互动率或网页观感自行补充接口未返回的风险结论。
- 检测失败或无权限时，不要切换网页搜索伪造结果。
推荐展示
达人：{creator_profile.nick_name}
平台：{platform}
粉丝：{creator_profile.followers_cnt}
假粉率估算：{fake_follower_rate * 100}%
互动质量分：{quality_score}/100
风险等级：{risk_level}
结论：{conclusion}
异常信号：{abnormal_types}
数据状态：完整 / 部分数据可用
signals 和受众画像只展示接口实际返回且有值的字段。
计费与权限
- 按
/openapi/v1/fake-follower-audit/run后台计费规则和租户套餐扣费，不在 Skill 内写死 credits。 - 仅在后端返回
40201时提示积分不足。 - 返回
40301时提示当前 API Key 未开通假粉检测权限，不要自动重试。 - 相同达人可能命中后端 7 天缓存，但是否扣费仍以后端响应和计费规则为准。