anthropics/skills180kwebapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
浏览器自动化
Check that a form actually rejects bad input: the error message renders, the submit button stays disabled, and no request goes out. Use when validation logic was written but never driven, when a "required" or pattern check looks right on screen but was never proven, or when a bug report says the form submitted invalid data anyway.
把这段话发给 Claude Code、Codex 或 Cursor。智能体会先检查安全性,你确认后才安装。
读取 https://funcoding.ai/skills/reticlehq/reticle/verify-form-validation/install.md ,按里面的步骤帮我安装这个 Skill。
Validation code is written once, glanced at in the browser with one obviously-bad value, and never driven again. The bug that ships is never the value you tried. It's the one you didn't: a submit button that isn't disabled until the field blurs, a button that's disabled only by CSS and still clickable, or a handler that fires the request before the check finishes.
Reticle can drive the real form and check all three failure points at once. Not installed? RETICLE_INSTALL_SOURCE=npx_skill npx @reticlehq/server@latest init, then the install-and-verify skill.
required, pattern, type="email" and friends stop the browser at the constraint-validation bubble before your app's JS ever runs. If you assert on that native tooltip, you've verified the browser, not the code you're supposed to be testing. Confirm the error you're checking is the app's own element (a testid, a role, a rendered string), not a :invalid pseudo-state, before you trust the verdict. If the form relies on native validation alone with no app-level check behind it, that's the finding: say so, don't paper over it with a predicate that happens to pass.
Name the consequence before you act, same as any Reticle drive:
reticle_look({ action: "page", sessionId, mode: "interactive" }) // get refs for the field and submit control
reticle_act_and_wait({ sessionId, ref, action: "fill", args: { value: "<invalid>" }, until: { kind: "allOf", predicates: [
{ kind: "element", query: { testid: "field-error" } },
{ kind: "element", query: { role: "button", name: "Submit" }, state: "disabled" },
]}})
If the check runs on blur rather than on keystroke, move focus to a different control as its own step before asserting: reticle_act({ sessionId, ref: otherRef, action: "focus" }) on the submit button or the next field, since focusing anything else is what fires the blur you need. If it's debounced instead, don't sleep for it. Use reticle_run({ tool: "reticle_clock", args: { sessionId, advanceMs } }) to advance past the debounce window exactly as in test-error-states. A fixed sleep passes on your machine and flakes in CI.
The button looking disabled is not the same claim as the request never leaving. Don't just trust the disabled state you already asserted. Try the submit anyway, so a fake-disabled control (CSS-only, still clickable) gets caught instead of waved through:
reticle_act({ sessionId, ref: submitRef, action: "click" })
Then assert the negative. Checking count: 0 the instant after the click proves nothing on its own: the request may simply not have been sent yet. Give the app a real window to have tried before you trust the absence, settling first and then checking the count, in one call:
reticle_assert({ sessionId, since, timeout_ms: 3000, predicate: { kind: "allOf", predicates: [
{ kind: "settled" },
{ kind: "net", method: "POST", urlContains: "/api/...", count: 0 },
]}})
The timeout_ms is what makes this a wait instead of a snapshot: reticle_assert defaults to timeout_ms: 0, one evaluation at the instant you call it, so without it settled is just checked once right after the click and proves nothing about whether the app was actually idle. With timeout_ms: 3000, the call keeps polling for up to three seconds until the page genuinely goes quiet. Only once it has settled does a count: 0 reading mean anything. Use since from the click's own result, not from the earlier fill, so you're scoped to requests after the submit attempt specifically.
Correct the value and confirm the rejection was conditional, not permanent:
reticle_act_and_wait({ sessionId, ref, action: "fill", args: { value: "<valid>" }, until: { kind: "allOf", predicates: [
{ kind: "element", query: { testid: "field-error" }, absent: true },
{ kind: "element", query: { role: "button", name: "Submit" }, state: "enabled" },
]}})
A form that never re-enables once it has rejected something once is a second bug wearing the first one's clothes.
Only verified: "yes" is a pass. "no" is a real finding, "unknown" means Reticle couldn't tell and needs a better capture, and "no-fault" means nothing was disproven but nothing was declared either. None of the three are evidence the form rejects bad input.
disabled state is real, not opacity/cursor styling that only looks inert. This is the same distinction design-system-compliance draws between disabled-looking and disabled.A verdict here is about this field and this invalid value, not the whole form. A required-field check tells you nothing about a pattern check on a different field. If you only drove one input, say which one, and say the rest of the form is unverified rather than letting a single pass read as "the form validates."
Capability reference: curl https://docs.reticle.sh/capabilities.md. Everything else: curl https://docs.reticle.sh/llms.txt.
anthropics/skills180kToolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
浏览器自动化
addyosmani/agent-skills103kTests in real browsers via Chrome DevTools MCP. Use when building or debugging anything that runs in a browser. Use when you need to inspect the DOM, capture console errors, analyze network requests, profile performance, or verify visual output with real runtime data. Requires the chrome-devtools MCP server to be configured.
浏览器自动化
ComposioHQ/awesome-claude-skills77kToolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
浏览器自动化
code-yeongyu/oh-my-openagent70kDrives a real browser through the omowright library from the js eval kernel: sites the user is already signed into, forms and clicks, JS-rendered pages, screenshots, web QA, extension popups, a human handoff for login, CAPTCHA or OTP, and a browser you own for scraping, bot-scored targets, network capture and QA traces. Use for any interactive browser task; not for a plain search or an unblocked static fetch.
浏览器自动化
shanraisshan/claude-code-best-practice67kBrowser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.
浏览器自动化
CherryHQ/cherry-studio52kRun Cherry Studio critical-path system regression tasks through the repository-owned Playwright E2E workflow. Use for full regression, release acceptance, development-branch system validation, or a named cherry-regression-test task on GitHub-hosted macOS and Windows runners.
浏览器自动化