The agent assembled a comprehensive overview of FORMEL SKIN by relying almost entirely on web search results and prior knowledge, since the primary website (steps [0] and [2]) returned only raw HTML that was not directly parseable for content. The agent successfully explained what they do (telemedicine dermatology with personalized compounding), who it's for (acne, hyperpigmentation, etc.), pricing (€49/month), and key differentiators (custom formulations vs. stock prescriptions). However, the site itself did not surface this information in an agent-friendly format—the agent had to reconstruct the narrative from external sources (Google Startups, KALTBLUT Magazine, Dealroom, app store listings) and acknowledge significant gaps (specific ingredients, insurance coverage, geographic limits, clinician qualifications, clinical evidence).

• ›The formelskin.de homepage (steps [0], [2]) returned a Next.js rendered page that was not semantically structured for extraction; the agent could not parse pricing, service description, or differentiators directly from the site itself.
• ›All substantive content came from external sources: Google for Startups profile ([1]), KALTBLUT Magazine feature ([8]), Dealroom acquisition news ([11]), and Google Play store listing ([12]). The site did not publish this information in a machine-readable or easily discoverable form.
• ›Attempts to navigate the site's own navigation (steps [3]–[4], /preise and /über-uns) returned 404 errors, suggesting the site uses client-side routing that does not surface structure to fetchers or that information architecture is flattened.
• ›The agent explicitly cited sources from the search results, indicating those snippets and linked articles contained the actual answers, not the primary website.
• ›The 'heavy bridge' friction outcome reflects the gap: the agent had to bridge from an inaccessible website to third-party sources (press, app stores, funding databases, blogs) to answer the task.