The agent assembled a comprehensive explanation of treasury.sh by combining direct site fetches with web search and prior knowledge, working around a JavaScript-heavy website that blocked full content access. The agent successfully answered what the product does, who it's for, pricing, and differentiation from competitors, but had to supplement thin or inaccessible on-site content with external sources (Twitter, LinkedIn, search results) and inferred context.

• ›Steps [1], [2], [3], [14] fetched the homepage, pricing, about, and learn pages but returned truncated HTML heads only—the agent could not access the rendered content because treasury.sh is built with SvelteKit (JavaScript-heavy). The full feature lists, pricing tiers, and detailed positioning were inaccessible via direct fetch.
• ›Step [9] search results included the treasury.sh homepage link but no enriched snippet; the agent had to rely on prior knowledge and Twitter/social sources (steps [10], [15]) to surface the 'Thinking Steps' feature, AI accuracy claims, and founder identity—core differentiators that should have been on the main site.
• ›Step [15] search results linked to LinkedIn and X/Twitter profiles, which became the agent's primary source for product positioning and features because the official site did not serve this information in a machine-readable form. The agent inferred competitive positioning (vs. YNAB, Monarch Money) from external sources, not from treasury.sh's own comparison.
• ›Pricing was confirmed from step [2] (likely visible in the rendered page despite truncated HTML), allowing the agent to state $95/year and no-tier model with confidence, but most feature details came from search context and prior knowledge, not from the site itself.
• ›The agent encountered a 404 on /docs (step [4]), forcing reliance on a /learn page (step [14]) and external sources rather than official documentation—low site discoverability for agent-relevant content.