Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add shengyy/agent-skills --skill app-verificationgit clone --depth 1 https://github.com/shengyy/agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/shengyy/agent-skills/app-verification)<a href="https://agentmods.dev/skills/shengyy/agent-skills/app-verification"><img src="https://agentmods.dev/badge/skills/shengyy/agent-skills/app-verification/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/shengyy/agent-skills/app-verification"><img src="https://agentmods.dev/badge/skills/shengyy/agent-skills/app-verification.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00113 | $0.01838 |
| Opus 5.5 | $0.00045 | $0.00735 |
| Sonnet 5.5 | $0.00023 | $0.00368 |
| Haiku 4.5 | $0.00011 | $0.00184 |
Grade A, and why
app-verification scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- **驱动**:agent 怎么程序化地操作它。先找现成的(Playwright/Cypress 用例、expect 脚本、PTY 助手、能 curl 的接口、调试端口),没有再选通用做法:Web 和 Electron 用浏览器自动化或 CDP,CLI/TUI 用 tmux 或 PTY,服务直接发 HTTP。 How it starts
The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
app-verification — 让任何 agent 都能把应用真正跑起来
编译通过、测试全绿、作者自述,都不等于改动在用户手里成立。验证常常是整件事里最慢的一步,因为它在等人去点一遍。把「启动应用、像用户一样操作、留下证据」写成项目本地的 skill,之后每个 agent 都能自己完成这一步:派工时的交付能写成可跑的检查,独立验证有了统一的驱动方式。
两种用法:
- 创建:项目还没有这样的 skill。按本文从「先问仓库」做到「交付前跑一遍」。
- 维护:已有 verify skill,应用变了、某次驱动因为文档过时而失败,或定期体检。读 references/maintain.md。
生成的 skill 写给下一个 agent:它在任务中途冷启动读到这份文件,从没见过这个应用。
先问仓库,不问用户
下面几件从代码里回答,只问用户看不到的:
- 表面:用户实际碰的是什么——Web UI、桌面应用、CLI/TUI、API、移动端、库。有多个就挑主要的,其余记下。
- 启动:本地怎么起,优先用仓库自己写明的开发命令;端口、环境变量、种子数据、鉴权。
- 驱动:agent 怎么程序化地操作它。先找现成的(Playwright/Cypress 用例、expect 脚本、PTY 助手、能 curl 的接口、调试端口),没有再选通用做法:Web 和 Electron 用浏览器自动化或 CDP,CLI/TUI 用 tmux 或 PTY,服务直接发 HTTP。
- 观察:能留下什么证据——截图、终端记录、响应体、日志、退出码、数据库状态。
- 隔离:两个实例能不能并排跑(端口、数据目录、profile)。不能就在生成的 skill 里写明:宁可拒绝驱动共享实例,也别弄坏用户正在用的会话。
仓库现状跑不起来,先修好或准确报告再生成:在坏基线上写出来的步骤会教错。和验证无关的缺失资源挡住启动时(API 不用的静态目录、示例配置),生成的 skill 可以补上,标明是验证脚手架,清理时删掉。
生成的 skill 长什么样
放在当前 agent 会自动加载的项目 skill 目录;项目里已有 skill 目录就放在一起。命名 verify-<app>,frontmatter 的 description 写明应用、表面和什么时候用——没有 frontmatter 就不会被识别。正文每一节都来自上面问到的实情,不留占位符:
- 启动:让应用进入可用状态的确切命令,以及怎么判断已就绪(某行日志、某个提示符)。要能从空数据目录起到可用,含迁移和种子数据;就绪指能用,不只是进程有应答——数据库健康检查可能在初始化脚本还没跑完时就通过。写上怎么关。短命的 CLI/TUI 没有常驻进程:启动指构建一次,每次驱动各开一个隔离的 PTY 或 tmux 会话。
- 体检:一条只读检查,回答「这个实例值不值得驱动」——进程在、版本对、端口归我们、鉴权有效。任何时候看着不对,先跑它。
- 驱动:用仓库里真实的选择器和命令,不是示例。优先稳定的句柄(ARIA 标签、data 属性、提示符、路由),不用坐标和 tab 顺序。驱动是给下一个 agent 验证它刚改的行为用的,所以要能单独走一个功能、换上它自己的输入,在隔离的状态里执行并留证。一键跑完全部旅程的回归脚本可以有,但代替不了单步驱动:它断言的是写脚本时的行为,新行为它看不见,照样报通过。
- 证据:要留什么、放在哪(不进版本控制:证据属于那一次运行,提交了就会过期),以及证据的标准:走真实的用户路径,不用内部 setter 或测试专用接口;记录动作和动作后的状态,不只是最后一屏;可见结果之外核对副作用(写出的文件、插入的行、发出的消息);只在生产环境本来就隔离外部系统的边界上用 mock。安全路径是 dry-run 或测试模式时,靠观察(文件、网络、git refs)确认它到底跳过了什么,别信名字。
- 清理:拆掉这次起的实例和临时状态。只杀自己起的进程,不按进程名杀。证据不在清理范围内,留在 skill 写明的位置。
- 附带脚本:可执行,调用方式写在正文里。要读者逆向才能用的脚本不算助手。
功能地图
在 skill 目录下建 features/README.md 作索引,每个面向用户的功能一个文件,先覆盖最重要的 3–5 个(从路由、命令、菜单或文档里找)。每个文件从用户视角回答:这是什么功能,用户怎么走到它,用驱动方式一步步怎么操作(每步的具体命令),看到什么最终状态算它正常,有哪些坑。写法参照 references/feature-map-example.md。
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 58 lines · 113 tokens per session scan A c2d1bd8220a9
app-verification is a skill published in the GitHub repository shengyy/agent-skills (2 stars, last pushed 3d ago), licensed MIT. It adds 113 tokens to every session and 1,838 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-04.
app-verification compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| app-verificationshengyy/agent-skills | Skill | 2 | 3d ago | MIT | A | 113 | Claude Code, Codex |
| browser-testing-with-devtoolsaddyosmani/agent-skills | Skill | 114k | 5d ago | MIT | B | 68 | Claude Code |
| actionbook-web-testactionbook/actionbook | Skill | 1.6k | 19d ago | Apache-2.0 | A | 71 | Claude Code |
| web-testNikolay-Shirokov/cc-1c-skills | Skill | 760 | yesterday | MIT | A | 44 | Claude Code |
| playwright-visual-testingmanagedcode/dotnet-skills | Skill | 508 | yesterday | MIT | A | 101 | Claude Code, Codex |
| appbuilder-e2e-testingadobe/skills | Skill | 319 | yesterday | Apache-2.0 | A | 144 | Claude Code |
| qa-systematicMathews-Tom/armory | Skill | 340 | 6d ago | MIT | A | 62 | Claude Code |