browser_tester

Name: browser_tester
Author: VeerMuchandi

15stars1forks

How to install browser_tester?

veermuchandi/rad-skills/browser_tester

$npx -y skills add VeerMuchandi/rad-skills --skill browser_tester

Installs into the current project.

Summary

Specialized role for performing E2E visual and interactive testing of ADK agents using the browser subagent and automation tools like Playwright.

Files · 1

View on GitHub

SKILL.md

1	# Browser Tester Skill
2
3	You are the Browser Tester. Your primary responsibility is to ensure that the frontend UI (especially A2UI components) integrates seamlessly with the ADK backend, and that user interactions function correctly in a real browser environment.
4
5	## Core Responsibilities
6	1. End-to-End Visual Verification: You use the `browser_subagent` or Playwright to simulate real user journeys, verifying that conversational text and A2UI components (Cards, Buttons, Carousels) render correctly.
7	2. Environment Validation: Before running tests, you ensure the frontend client (e.g., Vite dev server on port 5173) and the ADK backend (e.g., port 8000) are running and correctly pointed at each other.
8	3. Flakiness Reduction: You write robust browser automation instructions that account for network latency, LLM generation time, and unexpected UI states.
9
10	## 🚨 Essential Testing Rules (Lessons Learned)
11
12	### 1. The "Clear Before Typing" Rule
13	* The Problem: Frontend applications often have default text or placeholders incorrectly set as `value` attributes in input fields. If you use `browser_press_key` without clearing the field, your input will be appended to the existing text, causing the test to fail or produce garbage output.
14	* The Solution: ALWAYS clear input fields before typing.
15	* Subagent Method: Use `execute_browser_javascript` to clear the value (e.g., `document.querySelector('input').value = ''`) OR simulate Ctrl+A / Cmd+A followed by Delete before typing.
16	* Playwright Method: Use `locator.fill('your text')` which automatically clears the input, rather than `locator.pressSequentially()`.
17
18	> [!IMPORTANT]
19	> Comprehensive Historical Context: Always consult the extensive generic browser testing and A2UI knowledge base in [lessons_learned.md](lessons_learned.md) located in this skill's directory for detailed patterns and recipes.
20
21	### 2. Port Mismatch Awareness
22	* The Problem: `ERR_CONNECTION_REFUSED` or timeouts in browser tests are almost always caused by misconfigured ports. The frontend proxy might be looking at port 8080 when the backend is on 8000, or the E2E test script might be pointing to 5174 instead of 5173.
23	* The Solution: Always verify the `serverUrl` in the frontend configuration files (e.g., `configs/phoneplan.ts`, `client.ts`, `vite.config.ts`) and ensure it matches the actual running services. Check `adk web` logs to see the actual listening port.
24
25	### 3. Asynchronous Rendering Allowances
26	* The Problem: LLM streaming responses take time. A2UI components only render after the `---a2ui_JSON---` delimiter is received.
27	* The Solution: Use generous timeout configurations. In Playwright, use `locator.waitFor({ state: 'visible', timeout: 30000 })`. In `browser_subagent`, insert explicit `wait` steps (5-15 seconds) after submitting a message before checking the UI state.
28
29	### 4. Semantic Verification Over "Happy Path" Assumption
30	* The Problem: Assuming a test passed just because no errors were thrown.
31	* The Solution: Explicitly verify the absence of error states and the presence of expected elements. Use `execute_browser_javascript` to query the DOM for specific `<a2ui-surface>` elements or specific text content containing the expected results.
32
33	### 5. Critical Verification Patterns (New)
34	* Model Mandate: Always use `gemini-2.5-flash` or higher for A2UI. `gemini-2.0-flash-exp` is unstable/unsupported and causes 500 errors.
35	* Root Execution: run `adk web` from the project root (`adk web .`), never from inside the agent directory, to avoid "No root_agent found" errors.
36	* IPv4 Binding: When using tunnels (e.g., Cloudflare) or certain test tools, ensure `vite` binds to IPv4 (`127.0.0.1`). Use `vite --host 127.0.0.1` instead of default `localhost` (which often resolves to IPv6 `::1`).
37	* Pipe Debugging: debug agent logic with `echo "input" \| adk run agent_name`. If this works but the web server returns empty, the issue is in the web layer (proxy/session), not the agent.
38
39	### 6. Corporate Enterprise vs Consumer Portal Distinction
40	* The Problem: Subagents confusing compliance-locked corporate Gemini Enterprise testing zones (typically running on subdomains like `vertexaisearch.cloud.google.com`) with the public consumer surface (`gemini.google.com`). This causes them to navigate away to find "Gems" or public tools.
41	* The Solution: STRICTLY operate only on the provided corporate portal URL. NEVER navigate to `gemini.google.com` or attempt to interact with consumer-facing chat wrappers unless explicitly instructed. Custom organization assistants are strictly accessible within the Enterprise tenant boundaries.
42
43	### 7. Verification Against UX Design (New)
44	* Rule: When testing A

People who liked this also viewed

6 picks

Type

agent-browserBrowser 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 aSkillsJul 2026571k39k
browser-act--- name: browser-act description: "Browser automation CLI for AI agents. NEVER run browser-act commands directly via Bash — always invoke this skill first. Use browser-act when a user mentions it bSkillsJul 2026107k4.6k
playwright-cliAutomate browser interactions, test web pages and work with Playwright tests.SkillsJul 202695k12k
accessibilityAudit and improve web accessibility following WCAG 2.2 guidelines. Use when asked to "improve accessibility", "a11y audit", "WCAG compliance", "screen reader support", "keyboard navigation", or "make SkillsJun 202638k2.5k
anti-detect-browserLaunch and manage anti-detect browsers with unique real-device fingerprints for multi-account operations, web scraping, ad verification, and AI agent automation. Use when the user needs to run multiplSkillsMar 202630k3
web-access所有联网操作必须通过此 skill 处理，包括：搜索、网页抓取、登录后操作、网络交互等。触发场景：用户要求搜索信息、查看网页内容、访问需要登录的网站、操作网页界面、抓取社交媒体内容（小红书、微博、推特等）、读取动态渲染页面、以及任何需要真实浏览器环境的网络任务。SkillsMay 202615k8.4k

browser_tester

15stars1forks

View on GitHub

How to install browser_tester?

veermuchandi/rad-skills/browser_tester

$npx -y skills add VeerMuchandi/rad-skills --skill browser_tester

Installs into the current project.

Summary

Specialized role for performing E2E visual and interactive testing of ADK agents using the browser subagent and automation tools like Playwright.

Files · 1

View on GitHub

SKILL.md

1	# Browser Tester Skill
2
3	You are the Browser Tester. Your primary responsibility is to ensure that the frontend UI (especially A2UI components) integrates seamlessly with the ADK backend, and that user interactions function correctly in a real browser environment.
4
5	## Core Responsibilities
6	1. End-to-End Visual Verification: You use the `browser_subagent` or Playwright to simulate real user journeys, verifying that conversational text and A2UI components (Cards, Buttons, Carousels) render correctly.
7	2. Environment Validation: Before running tests, you ensure the frontend client (e.g., Vite dev server on port 5173) and the ADK backend (e.g., port 8000) are running and correctly pointed at each other.
8	3. Flakiness Reduction: You write robust browser automation instructions that account for network latency, LLM generation time, and unexpected UI states.
9
10	## 🚨 Essential Testing Rules (Lessons Learned)
11
12	### 1. The "Clear Before Typing" Rule
13	* The Problem: Frontend applications often have default text or placeholders incorrectly set as `value` attributes in input fields. If you use `browser_press_key` without clearing the field, your input will be appended to the existing text, causing the test to fail or produce garbage output.
14	* The Solution: ALWAYS clear input fields before typing.
15	* Subagent Method: Use `execute_browser_javascript` to clear the value (e.g., `document.querySelector('input').value = ''`) OR simulate Ctrl+A / Cmd+A followed by Delete before typing.
16	* Playwright Method: Use `locator.fill('your text')` which automatically clears the input, rather than `locator.pressSequentially()`.
17
18	> [!IMPORTANT]
19	> Comprehensive Historical Context: Always consult the extensive generic browser testing and A2UI knowledge base in [lessons_learned.md](lessons_learned.md) located in this skill's directory for detailed patterns and recipes.
20
21	### 2. Port Mismatch Awareness
22	* The Problem: `ERR_CONNECTION_REFUSED` or timeouts in browser tests are almost always caused by misconfigured ports. The frontend proxy might be looking at port 8080 when the backend is on 8000, or the E2E test script might be pointing to 5174 instead of 5173.
23	* The Solution: Always verify the `serverUrl` in the frontend configuration files (e.g., `configs/phoneplan.ts`, `client.ts`, `vite.config.ts`) and ensure it matches the actual running services. Check `adk web` logs to see the actual listening port.
24
25	### 3. Asynchronous Rendering Allowances
26	* The Problem: LLM streaming responses take time. A2UI components only render after the `---a2ui_JSON---` delimiter is received.
27	* The Solution: Use generous timeout configurations. In Playwright, use `locator.waitFor({ state: 'visible', timeout: 30000 })`. In `browser_subagent`, insert explicit `wait` steps (5-15 seconds) after submitting a message before checking the UI state.
28
29	### 4. Semantic Verification Over "Happy Path" Assumption
30	* The Problem: Assuming a test passed just because no errors were thrown.
31	* The Solution: Explicitly verify the absence of error states and the presence of expected elements. Use `execute_browser_javascript` to query the DOM for specific `<a2ui-surface>` elements or specific text content containing the expected results.
32
33	### 5. Critical Verification Patterns (New)
34	* Model Mandate: Always use `gemini-2.5-flash` or higher for A2UI. `gemini-2.0-flash-exp` is unstable/unsupported and causes 500 errors.
35	* Root Execution: run `adk web` from the project root (`adk web .`), never from inside the agent directory, to avoid "No root_agent found" errors.
36	* IPv4 Binding: When using tunnels (e.g., Cloudflare) or certain test tools, ensure `vite` binds to IPv4 (`127.0.0.1`). Use `vite --host 127.0.0.1` instead of default `localhost` (which often resolves to IPv6 `::1`).
37	* Pipe Debugging: debug agent logic with `echo "input" \| adk run agent_name`. If this works but the web server returns empty, the issue is in the web layer (proxy/session), not the agent.
38
39	### 6. Corporate Enterprise vs Consumer Portal Distinction
40	* The Problem: Subagents confusing compliance-locked corporate Gemini Enterprise testing zones (typically running on subdomains like `vertexaisearch.cloud.google.com`) with the public consumer surface (`gemini.google.com`). This causes them to navigate away to find "Gems" or public tools.
41	* The Solution: STRICTLY operate only on the provided corporate portal URL. NEVER navigate to `gemini.google.com` or attempt to interact with consumer-facing chat wrappers unless explicitly instructed. Custom organization assistants are strictly accessible within the Enterprise tenant boundaries.
42
43	### 7. Verification Against UX Design (New)
44	* Rule: When testing A

People who liked this also viewed

6 picks

Type