kernel-agent-browser_skill

This skill guides automating websites with Kernel cloud browsers, handling bot detection, login persistence, and session management for reliable browser tasks.

Shell

1

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill kernel/skills --skill kernel-agent-browser

SKILL.md15.2 KB

Overview

This skill documents best practices for using agent-browser with Kernel cloud browsers. It focuses on reliable login automation, session persistence, dealing with bot detection, iframes, and locating Kernel browser session IDs and live view URLs. Use it to build robust site-specific browser automation workflows with kernel as the provider.

How this skill works

The skill explains environment configuration, login workflows, snapshot-driven element discovery, and interaction patterns for agent-browser -p kernel. It shows how to persist profiles, capture element refs, run JS for tricky elements, and retrieve Kernel session IDs and live view URLs when manual intervention is needed. It also covers waiting strategies, iframe constraints, and cleanup steps.

When to use it

Automating website interactions using agent-browser with the Kernel provider (-p kernel).
When you need persistent cookies, saved logins, or profile-based session persistence.
If the site uses bot detection and you may need to use stealth mode or manual live-view login.
When workflows involve cross-origin iframes requiring a Playwright script against the Kernel session.
When you need to find Kernel browser session IDs or live view URLs for debugging or manual login.

Best practices

Set KERNEL_API_KEY and KERNEL_PROFILE_NAME before the first agent-browser -p kernel call so the CLI holds the correct state.
Enable KERNEL_STEALTH=true to reduce bot detection; increase KERNEL_TIMEOUT_SECONDS for longer interactive flows.
Always take a fresh snapshot before interacting; element refs change after navigation and DOM updates.
Use agent-browser eval with document selectors for stubborn elements or to bypass non-reactive controls.
If automated login fails, locate the live view URL via kernel browsers list and kernel browsers view <session id> and ask the user to complete login.

Example use cases

Create a site-specific skill that logs into an e-commerce account, navigates to order history, and scrapes recent orders.
Automate a two-step login portal: capture username page then password page and verify dashboard URL.
Persist a profile to maintain multi-step session state across runs (saved cookies and MFA tokens).
Handle a cross-origin iframe by retrieving the Kernel session ID and running a Playwright script against that session.
Recover from bot detection by switching to live-view manual login and resuming automated steps after verification.

FAQ

Set KERNEL_PROFILE_NAME before starting the browser; the profile is created and cookies are saved automatically when the session ends.

What if element refs like @e1 stop working?

Take a fresh snapshot before interacting. Refs change after navigation, DOM updates, and between sessions; document stable selectors when possible.

How can I get the live view URL for manual login?

Run kernel browsers list to find the most recent browser with your profile name, then kernel browsers view <session id> to retrieve the live view URL.