tekkiiiii/the-agency/agent-browser
Claude Code and agent skill by tekkiiiii/the-agency
- community
- ext:[{"type":"config","description":"Configuration overrides"}]
Checking that the SKILL.md file is still in the repository.
Use for headless browser automation and QA: \"automate browser\", \"QA test this site\", \"scrape webpage\", \"take screenshot\", \"automate login\", \"find bugs on this page\". Rust CLI with QA health scores, diff-aware regression, named session profiles and cloud providers. Not for the user's real Chrome (use /browser-harness).
Install
npx vskill@latest install tekkiiiii/the-agency/agent-browserSemi-extensibleMentions customization, but needs skill-specific knowledge. Read the extensible skills standard
SKILL.md
agent-browser
A fast native Rust CLI for headless browser automation, designed specifically for AI agents. No Playwright or Node.js required for the daemon.
Installation
On Linux, add --with-deps to install system dependencies.
Core AI Workflow (Always Use This Pattern)
Refs are the preferred selector method for AI. They come from snapshot output like:
Then use @e2, @e3 in subsequent commands.
Essential Commands
Navigation & Interaction
Snapshot & Screenshots
Get Info
Wait
Selectors
Use refs @eN from snapshot (preferred). Also supported:
- CSS:
"#id",".class","div > button" - Text:
"text=Submit" - XPath:
"xpath=//button" - Semantic:
agent-browser find role button click --name "Submit"
Sessions & Authentication
Command Chaining
Chain with && when you don't need intermediate output. Run separately when you need to parse output first.
JSON Output (for programmatic use)
Security Options
Diff & Debug
Cloud Browser Providers
When a local browser isn't available (serverless, CI/CD):
Tabs & Frames
Configuration File
Create agent-browser.json in project root:
Priority: CLI flags > env vars > project agent-browser.json > ~/.agent-browser/config.json
Default Timeout
Default is 25 seconds. Override:
Keep below 30000ms to avoid EAGAIN errors.
Structured QA Workflow (for browser-based testing)
When the user asks to "QA test", "find bugs", "run browser tests", or "test this site", follow this workflow.
QA Modes
| Mode | Scope | When to use | |
|---|---|---|---|
| Diff-aware | Changed files/routes | Default on feature branch — auto-detect from git diff | |
| Full | All pages | Systematic page-by-page | |
| Quick | Homepage + top 5 | 30-second smoke test | |
| Regression | Compare to baseline | Compare against prior baseline.json |
Diff-Aware Setup
QA Tiers
| Tier | Scope | Fixes | |
|---|---|---|---|
| Quick | Critical + High | 30-second smoke test | |
| Standard | + Medium | Systematic exploration | |
| Exhaustive | + Low/Cosmetic | Full depth |
QA Phases
Phase 1 — Initialize:
Phase 2 — Authenticate: Handle login. Try cookie import first:
If 2FA or CAPTCHA: hand off to user, then resume.
Phase 3 — Orient:
Detect framework type (SPA/Next.js/Rails/WordPress) from snapshot patterns.
Phase 4 — Explore: Per page:
Phase 5 — Document: For each issue found, document immediately with:
Report format:
Phase 6 — Health Score:
Compute health score across 8 weighted categories:
| Category | Weight | Scoring | |
|---|---|---|---|
| Console errors | 15% | 100 - (critical×25 + high×15 + medium×8 + low×3) | |
| Functional | 20% | Based on broken flows | |
| Accessibility | 15% | ARIA labels, keyboard nav, focus management | |
| UX | 15% | Navigation clarity, error messaging, loading states | |
| Visual | 10% | Layout issues, spacing, alignment | |
| Links | 10% | Broken links, missing hrefs | |
| Performance | 10% | Load times, network errors | |
| Content | 5% | Missing text, broken images |
Overall: weighted average. Report as N/100.
Phase 7 — Fix Loop:
For each issue, in severity order:
- Locate —
grepfor error messages, component names - Fix — minimal change only, no refactoring
- Commit — one per fix:
``bash git add -p git commit -m "fix(qa): ISSUE-NNN — short description" ``
- Re-test — before/after screenshots,
agent-browser diff snapshot - Classify — verified / best-effort / reverted / deferred
Stop conditions:
- 30+ fixes reached
- Fix risk exceeds 20% (might break something else)
- All Critical + High resolved
Phase 8 — Save Baseline:
Output Structure
QA Report Template
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/agent-browser/SKILL.md
Popularity
- Repository stars
- 3
- 7-day trend score
- 7
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Extension points
configConfiguration overrides
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/tekkiiiii/the-agency/agent-browser)