Protected pages and credentials may be exposed to the hosted browser and parallel workers
Source references: 5The Skill may ask for credentials for protected areas and uses hosted Firecrawl requests to browse, snapshot, or scrape pages. If the user authorizes login, real credentials and authenticated page content may be processed by a third party; parallel execution can widen the set of workers receiving that material.
Account login details, dashboard data, customer information, internal documents, or private page URLs could leave the user's local environment and enter Firecrawl's processing, logging, or output path.
The Skill requires a Firecrawl API key for hosted requests and instructs browsing, snapshots, and scraping, so page content reached during a run may be processed by Firecrawl. However, the claimed exposure of real login credentials is not clearly supported: it only suggests asking about “credentials/constraints” when blocked and expressly prohibits submitting real credentials unless the user explicitly authorizes it and has permission. Parallel execution is optional, with no instruction to share credentials or protected content with workers. A user can ask where login state, snapshots, and the API key are sent or retained, and restrict runs to public pages and one executor.
This assessment concerns the code and conditions shown, not proof that harm has occurred.inputs: - name: FIRECRAWL_API_KEY description: Firecrawl API key for hosted Firecrawl requests. required: true---Show 4 other places
Ask at most 1-3 concise questions only if blocked, such as the URL, desired flow focus, or credentials/constraints for protected areas.Use Firecrawl browser to open the product and navigate key flows. Snapshot at each step, scrape pages when useful, and document what the user sees and can do.Do not submit real credentials, purchases, or irreversible actions unless the user explicitly instructs and has permission.If appropriate, use sub-agents or equivalent parallel task runners:- Homepage and Marketing- Signup and Onboarding- Pricing and Plans- Docs and Developer Experience- Dashboard and Core Product- Help and SupportEach walker should return screens visited, actions taken, observations, friction, and source URLs.## Parallel WorkIf appropriate, use sub-agents or equivalent parallel task runners: