AGENT SKILL

helium-harness

Use Helium Harness for browser interaction in the Helium browser.

License 许可
MIT
最近更新
2026-08-17
Stars
★ 0
Forks
0

01 前言 Front Matter

name
helium-harness
description
Use Helium Harness for browser interaction in the Helium browser.

02 资源文件 14

03 SKILL.md 正文

Helium Harness

Helium Harness connects an agent to the user's Helium browser through the Chrome DevTools Protocol. Use it for interactive work that needs an existing Helium session, rendered page state, or browser UI actions.

Connection

Run a short Python script through the CLI. Helpers are pre-imported.

@'
ensure_real_tab()
print(page_info())
'@ | helium-harness

The first connection may open Helium's chrome://inspect/#remote-debugging page. Ask the user to enable remote debugging and accept Helium's prompt, then retry once they confirm.

For an unusual installation location, set HH_HELIUM_PATH or HELIUM_PATH to the executable before invoking the harness.

Workflow

  • Begin navigation with new_tab(url), then call wait_for_load().
  • Inspect accessibility state before interacting.
  • Prefer targeted DOM checks to screenshots unless visual layout matters.
  • After an interaction, collect the smallest state check that proves its result.
  • Stop at login, password, MFA, CAPTCHA, payment, or final form-submission steps unless the user takes over or explicitly confirms the exact action.

Examples

new_tab("https://example.com")
wait_for_load()
print(page_info())
tree = cdp("Accessibility.getFullAXTree")["nodes"]
print([node for node in tree if node.get("role", {}).get("value") == "button"])

Use BU_CDP_URL or BU_CDP_WS only when the user provides an explicit CDP endpoint. They override local Helium discovery.

04 同仓库更多 Skills