agent-browser skill: missing inspection & debugging commands

[tlvenn]

Summary

The agent-browser skill covers navigation, snapshots, interactions, get info, screenshots, and wait commands well, but omits several important command categories that are essential for debugging workflows.

Missing Commands

JavaScript Evaluation (eval)

agent-browser eval "document.title"
agent-browser eval "JSON.stringify(localStorage)"

Critical for inspecting client-side state, framework internals, and triggering events programmatically.

Console & Errors

agent-browser console          # View console.log, console.warn, etc.
agent-browser errors           # View unhandled JS exceptions, failed resources
agent-browser console --clear
agent-browser errors --clear

Essential for debugging JS hooks, interop issues, and failed asset loads.

Network Monitoring & Interception

agent-browser network requests                    # View captured requests
agent-browser network requests --filter "api"     # Filter requests
agent-browser network requests --clear
agent-browser network route "*/api/*" --abort     # Intercept/abort
agent-browser network route "*/api/*" --body '{}' # Mock response
agent-browser network unroute

Cookies & Storage

agent-browser cookies get|set|clear
agent-browser storage local|session

Browser Settings

agent-browser set viewport <w> <h>
agent-browser set device "iPhone 14"
agent-browser set geo <lat> <lng>
agent-browser set offline on|off
agent-browser set media dark|light|reduced-motion
agent-browser set headers '<json>'
agent-browser set credentials <user> <pass>

Element Debugging

agent-browser highlight <sel>
agent-browser get box <sel>
agent-browser get styles <sel>
agent-browser is visible|enabled|checked <sel>

Tracing & Recording

agent-browser trace start|stop [path]
agent-browser record start <path>
agent-browser record stop

Tabs

agent-browser tab list|new|close|<n>

Mouse

agent-browser mouse move|down|up|wheel
agent-browser drag <src> <dst>

Why This Matters

The current skill is great for page interaction (the snapshot + ref workflow), but an AI agent debugging a web app needs the inspection side too — especially eval, console, errors, and network. These are the commands that turn agent-browser from "automation tool" into "debugging tool."

All of these commands are already supported by agent-browser CLI — they just need to be documented in the skill.

[mvanhorn]

Opened PR #236 to address this. Adds documentation for all 9 missing command categories: JavaScript evaluation, console/errors, network monitoring, cookies/storage, device settings, element debugging, recording/tracing, tabs, and advanced mouse controls. All commands follow the existing formatting patterns in the skill.