skills.rest
vTRKA
vTRKACOMMUNITY
Explore all skills by this author →
GitHub

browser-agent-harness

Orchestrate bounded browser automation with evidence capture and PASS/PARTIAL/BLOCKED reports.

Updated Apr 27, 2026
Search Tags:#browser automation#devtools#playwright#evidence collection#test verification#sandbox safety#mcp discovery
One-click install
npx skills add https://github.com/vTRKA/supervibe --skill browser-agent-harness
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: browser-agent-harness
Source: https://github.com/vTRKA/supervibe/tree/main/skills/browser-agent-harness
Command: npx skills add https://github.com/vTRKA/supervibe --skill browser-agent-harness

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) and assets (resource) components.

What problem does it solve?

Browser agents often start unsafe, run without runtime capability proof, and leave no durable evidence for what happened during clicks, typing, navigation, and cleanup.

Core Features & Use Cases

  • Evidence-first browser harnessing: prepares a bounded browser automation packet that captures URL/route, interaction trace, console/network/DOM, screenshots, accessibility/performance signals, and cleanup proof before any completion claim.
  • Safety boundaries & anti-slop gates: enforces sandboxing, forbids sensitive/unsafe actions unless explicitly scoped, and supports design-facing gates that block approvals when required proof is missing.
  • Surface verification & handoff-ready reporting: uses MCP discovery and runtime availability checks (browser/Playwright/DevTools) to either run a harness-capable session or return PARTIAL/BLOCKED with a clear fallback.

Quick Start

Use the browser-agent-harness skill to create an evidence-backed browser session plan for your target page or local preview, then ask your AI tool to execute only the approved bounded actions and return the filled harness report.

Frequently Asked Questions about browser-agent-harness

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I capture browser automation evidence like network logs and screenshots in Playwright?▼

Browser automation evidence is captured by packaging live inspection, interaction traces, console and network logs, DOM states, and screenshots into a bounded harness report with cleanup receipts.

What is the best way to verify browser agent actions safely without sandbox risks?▼

Safe browser agent verification is enforced through safety boundaries that sandbox the runtime, forbid sensitive actions unless explicitly scoped, and block approvals when required proof is missing.

Can I use browser-agent-harness if Playwright or DevTools is not available on my host?▼

Runtime capability discovery checks for browser, Playwright, and DevTools availability, returning a PARTIAL or BLOCKED status with a clear fallback if the host environment lacks the required tools.

How do I generate a parseable report after running browser test interactions?▼

A parseable browserAgentHarnessReport is generated after execution, outputting PASS, PARTIAL, or BLOCKED status alongside required evidence lanes including URL routes, console logs, and accessibility signals.

Why does my browser automation lack proof of cleanup after completing page interactions?▼

Evidence-first browser harnessing mandates capturing cleanup proof before any completion claim, ensuring that navigation, clicks, and typing are paired with durable runtime receipts and safety checks.

When should I avoid running autonomous browser interactions without safety boundaries?▼

You should avoid autonomous browser interactions without safety boundaries when an agent must perform sensitive actions, requiring explicit scoping and design-facing gates to prevent unsafe execution.

Related Skills

Explore More Software Engineering Skills→
vTRKAvTRKA

mcp-discovery

Select MCP capabilities from live runtime registry evidence.

Community
Advanced
OrinksOrinks
24

ultrawork

Coordinate parallel independent tasks with evidence sharing and pass/fail criteria.

Community
Advanced
po4ykapo4yka
58

rust-unsafe

Audit unsafe Rust usage across RIPDPI crates for safety invariants.

Community
Advanced
shalevaminshalevamin

playwright

Automate real browser interactions from the terminal using playwright-cli.

Community
Intermediate