Skip to content

v0.9.0: browser-bridge audit + scaffold tools

Choose a tag to compare

@solomonneas solomonneas released this 23 May 19:51

First n8n-ops-mcp pieces designed to compose with browser-bridge end-to-end.

Added

  • n8n_audit_browser_bridge_usage — read-only scanner that walks every workflow and surfaces nodes calling the browser-bridge CLI (Execute Command, Code, SSH, legacy Function). Returns one finding per (workflowId, nodeName, platform, action) plus a per-platform action summary. Optional filters: platform, action, activeOnly, includeArchived. Paginates via cursor up to maxWorkflows (default 250, max 1000) with bounded-concurrency getWorkflow fan-out (default 3, max 8). Per-workflow fetch failures land in fetchErrors instead of failing the whole audit. Detection regex requires .js / .cjs / .mjs extension to avoid false positives from path mentions like cd /opt/browser-bridge.
  • n8n_scaffold_browser_bridge_node — pure local generator (no n8n API call) that emits a ready-to-paste n8n node calling browser-bridge <platform> <action>. Two modes: code-node (default; spawnSync with stdin JSON, surfaces exitCode + stderr) and execute-command (quoted <<'JSON' heredoc). Mirrors the patterns in browser-bridge's docs/n8n-usage.md. Validates platform and action as kebab slugs to keep them safe to interpolate into shell commands. Warns when execute-command mode is paired with non-empty input (heredoc bakes input in; no per-item wiring).

Why

The audit + scaffold pair answer two real questions that were previously a manual grep job: "where am I calling Linktree sync from?" and "how do I add a CoderLegion scan-comments step without copy-pasting the spawn shape from docs every time?"

Both tools are read-only / local-only — registered before the enableEdit gate. Full suite green at 130/130.