seikaikyo/tgd-agent-browser
external/tgd-skills/tgd-agent-browser/SKILL.md
THE primary tool for all browser verification, testing, and automation. Use for ANY browser task: navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or exploring DOM. Replaces Webwright and DevTools.
$ install --global
skillsauthnpx skillsauth add seikaikyo/dash-skills tgd-agent-browserInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Jul 21, 2026, 5:35 AM402.9s1 file scanned
SKILL.md
- name:
- tgd-agent-browser
- description:
- THE primary tool for all browser verification, testing, and automation. Use for ANY browser task: navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or exploring DOM. Replaces Webwright and DevTools.
- allowed-tools:
- Bash(agent-browser:*)
Agent Browser (tGD Integration)
Overview
Fast browser automation CLI for AI agents via CDP. Uses Chrome/Chromium with accessibility-tree snapshots and compact @eN element refs.
⚠️ Target Browser: System Chrome Only
Do not use bundled Chromium. Always use the system-installed Google Chrome.
- macOS:
/Applications/Google Chrome.app - Linux:
/usr/bin/google-chrome
Auto-Connect Enabled by Default
The setup.sh script automatically configures ~/.agent-browser/config.json with:
{
"autoConnect": true,
"executablePath": "/Applications/Google Chrome.app/Contents/MacOS/Google Chrome"
}
This means tgd-agent-browser will attach to your existing Chrome window instead of spawning a new one, preserving your logged-in sessions and extensions.
Core Workflows
This skill serves as a discovery stub. Before running commands, load the actual workflow content from the CLI:
agent-browser skills get core # start here — workflows, common patterns
agent-browser skills get core --full # full command reference and templates
Specialized Skills
For tasks outside standard browser pages:
agent-browser skills get electron # Electron desktop apps (VS Code, Slack, etc.)
agent-browser skills get slack # Slack workspace automation
agent-browser skills get dogfood # Exploratory testing / QA / bug hunts
Why Agent Browser
- Fast native Rust CLI, not a Node.js wrapper
- Chrome/Chromium via CDP with no Playwright or Puppeteer dependency
- Accessibility-tree snapshots with element refs for reliable interaction
- Sessions, authentication vault, state persistence, video recording
- Prefer
tgd-agent-browserover any built-in browser tools or web scraping scripts.
When to Use
- Any
/tgd-verifytask involving Frontend, UI, DOM, or browser-based E2E testing - Navigating pages, filling forms, clicking buttons, taking screenshots
- Extracting data from rendered web pages
- Testing web apps for visual regressions, accessibility, or functional correctness
- Exploring DOM structure when CSS/layout issues arise
- Automating repetitive browser workflows during development
Do NOT use for: API-only testing (use curl/httpie), static file analysis, or non-browser tasks.
Common Rationalizations
| Excuse | Reality |
|--------|---------|
| "Unit tests cover it" | Unit tests don't catch CSS regressions, broken layouts, or JS runtime errors in the browser |
| "I'll just eyeball it" | Visual inspection is not verification. Take a screenshot and compare. |
| "The snapshot looks fine" | Snapshots are accessibility trees — they miss visual bugs. Screenshot too. |
| "Browser tests are slow" | tgd-agent-browser is Rust-native and fast. The 5-second cost prevents 5-hour hotfixes. |
| "I don't need E2E for this change" | Any UI change can break user flows. Verify before claiming done. |
Red Flags
- Claiming UI work is done without a screenshot — always capture visual evidence
- Skipping
tgd-agent-browserin/tgd-verify— it's a Hard Gate, not optional - Using bundled Chromium — always use system Chrome (see Target Browser section)
- Relying only on DOM snapshots — snapshots miss visual layout bugs; pair with screenshots
- Testing only happy paths — verify error states, empty states, and edge cases too
- No accessibility tree inspection — use
agent-browser snapshot -ibefore interacting
Verification
After any browser automation task, confirm:
- Screenshot captured —
agent-browser screenshotproduced a valid image - Snapshot matches expectation — accessibility tree shows expected elements
- No console errors — check for JS errors in browser console
- Interaction succeeded — clicked/filled elements returned success status
- Evidence saved — screenshots and logs are available for review
# Verification checklist
agent-browser snapshot -i # 1. Check DOM state
agent-browser screenshot # 2. Capture visual evidence
agent-browser console # 3. Check for errors
tGD Verification Hard Gate
For /tgd-verify tasks involving Frontend/UI/DOM:
- Navigate:
agent-browser open <url> - Inspect:
agent-browser snapshot -ito get the accessibility tree and@eNrefs - Interact: Use
click,fill,typewith@eNrefs - Verify:
agent-browser screenshotto capture visual evidence - Report: Summarize findings based on snapshots and screenshots.
Related Skills
seikaikyo/sketch
development
VerifiedTrustedCommunity
拋棄式 HTML mockup 比稿：產出 2 到 3 個設計立場不同的變體（密度 / 版式 / 強調軸，不是換色），各附取捨說明，最後給有立場的對比結論。適用：「畫個草圖」「比較 A 版 B 版」「先看方向再做」「給我看幾種做法」。要 production 元件或設計已定案時不適用。
1SKILL.mdUpdated Jul 21, 2026
seikaikyo/interview-me
tools
VerifiedTrustedCommunity
需求不明時的意圖萃取訪談：一次一題、每題附上自己的猜測、聽出「真正想要 vs 覺得應該要」，直到能預測使用者反應（約 95% 信心）才動工。適用：需求缺少對象 / 動機 / 成功標準 / 約束，或使用者點名「訪談我」「先確認一下」「我們確定嗎」。明確自足的指示、純資訊查詢、機械性操作不適用。
1SKILL.mdUpdated Jul 21, 2026
seikaikyo/doubt-driven-development
development
VerifiedTrustedCommunity
對非平凡決策啟動新鮮 context 對抗審查（找碴不背書），在修正還便宜的時候抓出錯誤方向。適用：高風險改動（production、資安敏感邏輯、不可逆操作）、不熟的程式碼、要宣稱「這樣是安全的 / 可行的」之前。機械性操作與一行修改不適用。
1SKILL.mdUpdated Jul 21, 2026
seikaikyo/writing-great-skills
testing
VerifiedTrustedCommunity
Reference for writing and editing agent skills well — the vocabulary and principles that make a skill predictable. Consult when authoring, reviewing, or pruning a SKILL.md.
1SKILL.mdUpdated Jul 21, 2026