The agent successfully synthesized a usable explanation of agentkit.md from the site's homepage alone, but had to supplement with external search results to address the competitive positioning question the site itself didn't answer. The site publishes its core facts clearly but narrowly—it doesn't contextualize itself within the broader agent ecosystem.

• ›Steps [1–4] all fetched the same URL (agentkit.md/) yet returned distinct content sections (overview, pricing, target users, API endpoints), indicating the agent reused or synthesized information from a single comprehensive homepage response rather than discovering multiple pages.
• ›The site provides no explicit comparison to alternatives; the agent was forced to web-search [step 6] to answer 'how it's different,' filling a gap the domain did not address. This search revealed competitors (MuleSoft, Google Cloud, LangChain Hub, Agentregistry) the site never mentions.
• ›Core facts (free pricing, REST API, MCP server, Herdr harness focus, keyless access, AI-generated cards) all came from the homepage; the site publishes these as natural prose, not in machine-readable structured data (no JSON-LD, OpenAPI spec, or schema.org markup visible in fetches).
• ›The agent correctly flagged 'Herdr harness' as unexplained on the site itself—the domain assumes domain knowledge (what Herdr is) that new users lack, limiting discoverability for outsiders.
• ›Navigation was minimal: the agent typed the homepage URL from memory and followed no links; 75% of discovered URLs were said to follow links, but the trajectory shows only homepage fetches, suggesting either link discovery failed or the homepage is comprehensive enough that links were unnecessary.