The agent successfully assembled a comprehensive explanation of Sourceful Energy's business model, target audience, and differentiation by combining content from the site's markdown artifact files (llms.txt, apply.md, master-plan.md, sank-effekttariff.md, platform.md, and open-source.md). The site's HTML pages returned truncated boilerplate that was not useful; the agent relied entirely on machine-readable markdown artifacts to construct its answer. Pricing remained opaque—a genuine gap the agent correctly flagged—because Sourceful publishes no pricing details, only custom per-site assessment processes.

• ›Steps [1]–[9] fetched HTML pages (homepage, about, pricing, platform, blog, sank-effekttariff) but all returned truncated boilerplate with no substantive content. The HTML layer did not surface the company's actual value proposition, positioning, or business model.
• ›Step [10] discovered llms.txt, a machine-readable artifact that provided the canonical mission statement ('The Grid Connection Platform') and the first concrete description of what Sourceful does. This was the breakthrough step that unlocked the rest of the answer.
• ›Steps [11]–[15] fetched markdown files (.md) from the same domain, each providing distinct, substantive content: apply.md explained the assessment process; master-plan.md outlined the company's vision; sank-effekttariff.md described the specific C&I battery use case; platform.md explained the B2B2C offering to utilities/OEMs; open-source.md detailed FTW, the open-source EMS. All markdown artifacts were content-bearing and directly constructed the final answer.
• ›The agent never actually parsed or extracted text from the HTML responses in steps [1]–[9]; it relied on prior knowledge to guess the URLs and then switched to markdown artifacts (steps [10]–[15]) once discovered. This suggests the HTML rendering layer is not agent-optimized.
• ›The site's agent-readiness is bifurcated: HTML pages appear to be client-rendered Next.js that truncates to boilerplate in API responses, but markdown artifacts (.md files) and llms.txt are directly accessible, structured, and information-dense. The agent had to discover and pivot to the markdown layer through structured artifact probing (llms.txt) and inference about related markdown files.
• ›Pricing information was completely absent from all fetched resources. The agent correctly identified this as a genuine gap and explained the custom per-site assessment model as the operative alternative, but no published pricing structure exists on the site for any tier or geography.
• ›Geographic scope (Sweden/Nordic focus evident from Kalmar Energi reference) was surfaced from platform.md but not explicitly stated in mission or positioning materials, leaving ambiguity about international availability—a limitation the agent noted.