R
rey0905/paywall-radar
An agent skill that spots hidden paywalls before your AI recommends a site, without signing up.…
无 SECURITY.md
在 GitHub 查看 ↗链接已于 2026-11-12 核查 · HTTP 200 · 打开的是厂商自己的官网
安装与配置
客户端配置（取自该仓库的安装命令）
{
"mcpServers": {
"rey0905-paywall-radar": {
"command": "npx",
"args": ["skills"]
}
}
}解析自我们于 2026-10-26 抓取的该仓库 README；我们未执行过。
command/args 来自仓库写明的安装命令；不确定时请以该项目 README 为准。
VS Code（.vscode/mcp.json —— 键名是 servers）
{
"servers": {
"rey0905-paywall-radar": {
"command": "npx",
"args": ["skills"]
}
}
}VS Code 顶层键是 <code>servers</code>，其余与上面那份相同。
Claude Code（命令行）
claude mcp add --transport stdio rey0905-paywall-radar -- npx skills
Claude Code 不收 JSON 文件，用这条命令直接加。
配置项的键名以你的客户端为准：客户端与 CLI
- ★ 数取值
- 1 ★ · rey0905/paywall-radar · 取自 GitHub 搜索 API · （2026-11-13 20:00:50 核对）
- 来源明细
- 从该仓库的 GitHub 页面直接读取 · 条目类型：开源项目
- 收录与活跃
- 收录于 2026-10-21 · 最近见到 2026-11-14 · 共见到 28 次 · 被 1 个来源列出
- 评分构成
- 综合分 103 · 1 ★ · 1 个来源 · repo
- 安装
npx skills
常见问题（答案都来自我们的记录）
怎么安装？
该仓库给出的安装命令是
npx skills。我们的记录里它是仓库自己提供的；不保证在你的客户端上直接可用。它适合谁 / 用来做什么？
我们把它归在这些标签下: Agent 技能。标签来自条目自身，不是我们强加的分类。
谁维护它、还在更新吗？
仓库
rey0905/paywall-radar；我们收录于 2026-10-21；最近见到 2026-11-14。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。# paywall-radar
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
Stop your AI from recommending sites that make you fill in a 10-step quiz, then ask for your credit card.
An agent skill that screens websites for hidden paywalls before recommending them, without signing up or giving away any of your data.
English · 简体中文
## The problem
You ask your AI for remote job sites. It gives you a list. You click one, and it asks what role you want. Then your experience level, salary range and preferred time zone, then it wants your resume. Twenty minutes later:
> Choose your plan · ⏱ 09:52 · ~~£0.43~~ £0.07 per day · 50% DISCOUNT
> Without cancellation ... will automatically charge 155.99 GBP every 1 year until I cancel.
That one is real (observed 2026-10-22). The homepage never mentions a price, never says "free", and has no pricing page. Your AI had no way of knowing, so it recommended the site anyway.
paywall-radar teaches your agent to check first.
## What it does
Whenever your agent is about to recommend websites, apps or online tools, or you ask "is this site free?", it screens each one and sorts it:
| Verdict | Meaning |
| ✅ FREE | The thing you came to do is free |
| ✅ FREE TIER OK | The free tier gets the job done; paid extras are listed |
| ❔ UNVERIFIED | Not enough evidence, so it is marked "may charge" and not passed off as free |
| ⛔ FILTERED | You can't do the core task without paying, and it says whether the price is hidden or shown openly |
Filtered sites still get one line at the end with the reason, so you recognize them if you come across them elsewhere.
## How it sees the paywall without signing up
The hard part: the paywall sits behind the quiz, and the agent can't (and shouldn't) fill the quiz in with your details. So it looks around the funnel instead of through it:
1. Public info outside the funnel. It reads the pricing page, the FAQ, and the terms (searching for auto-renew, recurring, trial, refund). It checks app store in-app purchases and Reddit/Trustpilot reviews. A money-back policy on a site that never shows a price says a lot.
2. The site's own code. Every visitor's browser downloads the site's JavaScript. If that code contains routes like choosePlan, limitedOffer or /start/payment-process, there is a plan screen at the end of the flow, whatever the homepage says.
3. Skipping ahead by URL. Funnels often number their steps (/journey/2/join/10). Changing the number sometimes lands straight on the plan page.
4. Testing the real task. It opens a job listing and checks whether "Apply" works without an account, or whether the resume export is actually free.
5. No evidence of free means not free. Genuinely free products say "free" loudly, because it's their best selling point.
The bundled probe covers the code and policy-page checks in about 7 seconds per site:
text
$ python3 scripts/probe.py https://globalwork.ai/en
=== https://globalwork.ai/en
pricing link on page: no
says 'free' out loud: no
payment provider in code: none found
paywall/funnel code signals: choose-plan step, checkout / payment step, upsell / limited offer,
subscription API, pricing API, auto-renew wording
billing terms in policy pages: YES
[choose-plan step] choosePlan:"/platform/settings/manage-subscription/choose-plan",limitedOffer:...
An agent skill that screens websites for hidden paywalls before recommending them, without signing up or giving away any of your data.
English · 简体中文
## The problem
You ask your AI for remote job sites. It gives you a list. You click one, and it asks what role you want. Then your experience level, salary range and preferred time zone, then it wants your resume. Twenty minutes later:
> Choose your plan · ⏱ 09:52 · ~~£0.43~~ £0.07 per day · 50% DISCOUNT
> Without cancellation ... will automatically charge 155.99 GBP every 1 year until I cancel.
That one is real (observed 2026-10-22). The homepage never mentions a price, never says "free", and has no pricing page. Your AI had no way of knowing, so it recommended the site anyway.
paywall-radar teaches your agent to check first.
## What it does
Whenever your agent is about to recommend websites, apps or online tools, or you ask "is this site free?", it screens each one and sorts it:
| Verdict | Meaning |
| ✅ FREE | The thing you came to do is free |
| ✅ FREE TIER OK | The free tier gets the job done; paid extras are listed |
| ❔ UNVERIFIED | Not enough evidence, so it is marked "may charge" and not passed off as free |
| ⛔ FILTERED | You can't do the core task without paying, and it says whether the price is hidden or shown openly |
Filtered sites still get one line at the end with the reason, so you recognize them if you come across them elsewhere.
## How it sees the paywall without signing up
The hard part: the paywall sits behind the quiz, and the agent can't (and shouldn't) fill the quiz in with your details. So it looks around the funnel instead of through it:
1. Public info outside the funnel. It reads the pricing page, the FAQ, and the terms (searching for auto-renew, recurring, trial, refund). It checks app store in-app purchases and Reddit/Trustpilot reviews. A money-back policy on a site that never shows a price says a lot.
2. The site's own code. Every visitor's browser downloads the site's JavaScript. If that code contains routes like choosePlan, limitedOffer or /start/payment-process, there is a plan screen at the end of the flow, whatever the homepage says.
3. Skipping ahead by URL. Funnels often number their steps (/journey/2/join/10). Changing the number sometimes lands straight on the plan page.
4. Testing the real task. It opens a job listing and checks whether "Apply" works without an account, or whether the resume export is actually free.
5. No evidence of free means not free. Genuinely free products say "free" loudly, because it's their best selling point.
The bundled probe covers the code and policy-page checks in about 7 seconds per site:
text
$ python3 scripts/probe.py https://globalwork.ai/en
=== https://globalwork.ai/en
pricing link on page: no
says 'free' out loud: no
payment provider in code: none found
paywall/funnel code signals: choose-plan step, checkout / payment step, upsell / limited offer,
subscription API, pricing API, auto-renew wording
billing terms in policy pages: YES
[choose-plan step] choosePlan:"/platform/settings/manage-subscription/choose-plan",limitedOffer:...
收录徽章（可选）
想把收录标记放进 README 就复制这行，可选，不影响收录。
[](https://anclum.com/zh/s/rey0905-paywall-radar/)
- 来源
- github · https://github.com/rey0905/paywall-radar
- 按领域浏览
- Agent 技能
- Traction
- 1 ★
- 收录于
- 今天
- 安全策略
- 我们没在该仓库根目录找到
SECURITY.md（这不代表不安全）
同类推荐
所属领域与标签
每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。