The agent successfully answered all core questions about shadertown.com by synthesizing information from the homepage, docs, llms.txt, and OpenAPI schema. The site is moderately agent-ready: it publishes key facts in machine-readable form (llms.txt, OpenAPI) and has a docs section, but pricing and product details require parsing JavaScript-heavy pages or relying on structured metadata rather than plain HTML.

• ›Step [5] (llms.txt) was the single most valuable resource—it directly stated what shadertown is ('45 production WebGPU shaders'), who it's for, and pricing tiers ($4.99/month, $19.99 one-time, 13 free). The agent cited this as a source.
• ›Step [6] (OpenAPI schema) confirmed the same core facts and added technical details about the API and licensing model (402 response for unpaid shaders). The agent cited this as a source.
• ›Step [1] (homepage) and Step [4] (docs) were fetched but the agent had to rely heavily on prior knowledge and the llms.txt/OpenAPI outputs rather than extracting from the rendered HTML—the agent notes that the pricing page was 'JavaScript-heavy' and 'didn't fully render.'
• ›The site publishes detailed, structured metadata (llms.txt, OpenAPI) that is designed for machine consumption, making it unusually transparent to agents. However, the main marketing pages (homepage, pricing, shaders gallery) are client-rendered and don't yield their content in fetched HTML.
• ›The agent correctly identified gaps: specific free vs. paid shader breakdown, subscription/payment flow details, and account management mechanics were not clearly surfaced. These represent real information gaps, not just rendering failures.