Bundle Pluginsource linked

n8n Harnessv0.1.0

OpenClaw skill plugin for BrowserOS MCP automation through n8n workflows.

@khanh-atom/n8n-harness·runtime n8n-harness·by @khanh-atom

openclaw bundles install clawhub:@khanh-atom/n8n-harness

Latest release: v0.1.0Download zip

Capabilities

Bundle format

generic

Compatibility

Built With Open Claw Version

2026.4.27

Plugin Api Range

>=2026.4.25

Security Scan

VirusTotal

Benign

View report →

OpenClaw

Benign

high confidence

ℹ

Purpose & Capability

The included files align with the stated purpose: workflows open pages, click/fill elements, take snapshots, extract page content, and run Google/Google Calendar browser workflows through BrowserOS MCP. The main metadata mismatch is that the registry declares no required binaries, while the documentation clearly requires BrowserOS MCP, n8n-atom-cli, npm/npx, and mcp-remote to be available.

ℹ

Instruction Scope

The instructions stay within browser automation and workflow authoring. They do grant broad browser-control capability, including reading page content from authenticated sites such as Google Calendar if the user is logged in. One workflow performs an informational list_pages call before Google search, which is not strictly needed but remains local and is not sent to an unexpected endpoint.

ℹ

Install Mechanism

There is no OpenClaw install spec, so the skill itself does not install files or download archives. However, runtime instructions use npm exec n8n-atom-cli and npx mcp-remote, which may execute npm-provided tools if they are not already installed. This is coherent with the n8n/BrowserOS design but should be understood by the user.

✓

Credentials

The skill requests no environment variables, API keys, tokens, passwords, or config paths. It relies on the user's existing browser session for sites like Google Calendar, which is expected for browser automation and is documented.

✓

Persistence & Privilege

The skill is not marked always-on, does not activate on startup, and the TypeScript plugin entry is a no-op. It does not modify other skills or system-wide agent settings.

Assessment

Install this only if you want OpenClaw to automate your browser through BrowserOS MCP. The workflows can open pages, click, type, and extract page content, including content from sites where your browser is already logged in. There is no evidence here of unrelated credential access or unexpected external exfiltration, but you should make sure BrowserOS MCP is local/trusted and that n8n-atom-cli and mcp-remote come from sources you trust.

Verification

Tier

source linked

Scope

artifact only

Summary

Validated package structure and linked the release to source metadata.

Source

github.com/khanh-atom/n8n-atom-harness

Commit

8013c2ff1053

Tag

master

Provenance

Scan status

clean

n8n-harness

Browser automation via BrowserOS MCP and n8n workflow files.

No Python, no CDP daemon, no framework — just .n8n workflow files that make HTTP requests to BrowserOS MCP.

OpenClaw plugin

This repository is also an OpenClaw skill plugin:

openclaw.plugin.json declares the n8n-harness plugin.
index.ts is the OpenClaw runtime entry. It is intentionally a no-op because the plugin ships skills, not runtime tools.
SKILL.md is exposed to OpenClaw via the manifest skills: ["./"].

For local development, link this repo into OpenClaw:

openclaw plugins install -l .
openclaw plugins inspect n8n-harness

If this repo is placed under a workspace .openclaw/extensions/ directory instead, enable it explicitly:

openclaw plugins enable n8n-harness

Architecture

┌──────────────────────────┐
│ .n8n workflow file       │
│ (HTTP Request nodes)     │
└──────────┬───────────────┘
           │ POST /execute-tool
           ▼
┌──────────────────────────┐
│ BrowserOS MCP Server     │
│ localhost:6277            │
└──────────┬───────────────┘
           │
           ▼
┌──────────────────────────┐
│ Browser (full control)   │
└──────────────────────────┘

Quick start

Ensure BrowserOS MCP is running on localhost:6277
Run a skill: npm exec n8n-atom-cli run skills/browseros/open-page.n8n
Or call MCP directly:

curl -X POST http://localhost:6277/execute-tool \
  -H "Origin: http://localhost:6274" \
  -H "Content-Type: application/json" \
  -d '{
    "toolName": "new_page",
    "toolArgs": { "url": "https://www.google.com", "hidden": false, "background": false },
    "server": { "command": "npx", "args": ["mcp-remote", "http://127.0.0.1:9001/mcp"] }
  }'

Skills

Pre-built .n8n workflow files in skills/:

Skill	Description
`browseros-open-page.n8n`	Open a new browser page
`browseros-take-snapshot.n8n`	Get interactive element IDs
`browseros-click-element.n8n`	Click an element
`browseros-fill-input.n8n`	Type into an input field
`browseros-get-page-content.n8n`	Extract page content as markdown

See SKILL.md for full documentation, all available MCP tools, and how to write new skills.

How to create new skills

Use BrowserOS MCP directly to accomplish your task
Once you have the working flow, save it as a .n8n workflow file in skills/
Name it <site>-<action>.n8n for site-specific skills

License

MIT