Parallel
WarnAudited by ClawScan on May 10, 2026.
Overview
This is mostly a normal Parallel.ai research skill, but it embeds a fallback API key and can forward an undeclared BrowserUse credential to an external MCP service.
Before installing, remove the hardcoded Parallel key and use only your own PARALLEL_API_KEY. Keep BROWSERUSE_API_KEY unset unless you intentionally want authenticated browsing, and create monitors or webhooks only when you explicitly want persistent tracking.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If the environment variable is absent, searches could be sent using an unknown exposed account key, affecting privacy, billing/quota attribution, and exposing the embedded credential to anyone with the skill.
The main search script falls back to an embedded token instead of requiring the user's declared PARALLEL_API_KEY.
API_KEY = os.environ.get("PARALLEL_API_KEY", "y2s_m4er5i6-5qCikOLUtmnkvOYRU24eDphq_jg1")Remove the hardcoded fallback, require PARALLEL_API_KEY for all scripts, and rotate or revoke the exposed key.
A BrowserUse API key in the user's environment may be used during task runs even if the user did not pass it for that run, enabling authenticated browsing through another provider.
When BROWSERUSE_API_KEY is present, the task script attaches a BrowserUse MCP server and bearer token to Parallel task requests; this optional credential and inter-provider data flow are not declared in the main skill requirements.
browseruse_key = args.browseruse_key or os.environ.get("BROWSERUSE_API_KEY") ... "url": "https://api.browser-use.com/mcp" ... "headers": {"Authorization": f"Bearer {browseruse_key}"}Declare this optional credential and data flow, require an explicit flag or confirmation before adding MCP servers, and only include the key for tasks that need authenticated browsing.
If invoked, a monitor can keep running in the user's Parallel account and may send events to the configured webhook until deleted.
The monitor helper can create remote recurring monitors and webhook notifications, which may continue after the immediate chat task.
Parallel.ai Monitor API - Continuous web tracking with alerts. ... python3 monitor.py create "Alert when AirPods drop below $150" --cadence hourly --webhook https://...
Create monitors only on explicit user request, document them prominently, and list/delete monitors when no longer needed.
