The agent successfully assembled a comprehensive explanation of pen.dev by relying almost entirely on external review articles and web search results, since the site itself (pen.dev) returned only raw HTML with minimal readable content. The agent filled gaps using prior knowledge and third-party sources, ultimately delivering a complete answer that covered what the product does, its audience, pricing, and differentiation—but only because those sources existed; the site's own pages were not agent-ready.

• ›The pen.dev homepage, pricing page, and docs page [1, 3, 4, 9] all returned heavy Next.js JavaScript bundles with no server-rendered content readable in the HTML payload. The agent could not extract any substantive information from the official site itself.
• ›All useful content in the final answer came from external sources: the FlowStep pricing review [13], Medium articles and design comparisons indexed in web searches [7, 11, 15, 19]. These third-party sources provided the feature descriptions, pricing details, use case positioning, and competitor comparisons that the official site did not expose.
• ›The agent's answer explicitly cites seven sources, only three of which are pen.dev's own properties (pen.dev/, pen.dev/pricing, docs.pencil.dev/); the remaining four are external reviews and comparisons. This indicates the site publishes minimal machine-readable or crawl-friendly documentation.
• ›Key gaps the agent could not resolve from the site: the vague pricing roadmap, the name ambiguity (pen.dev vs. Pencil.dev), the exact scope of features, and whether Figma is a prerequisite or optional. These were acknowledged as confusing precisely because the official site did not address them directly.
• ›The agent worked around the inaccessibility of the site's own pages by pivoting to web search and prior knowledge (50% + 25% of fetch sources), a heavy-bridging pattern typical when a site relies on client-side rendering without server-side content exposure.