The agent retrieved BoosterLab's two main offerings (AI agent service and training courses) from three site pages and synthesized a coherent summary of their business model, audience, and pricing. The site is navigable but deliberately vague on pricing details and training content; the agent correctly identified gaps and flagged them rather than filling them from model memory.

• ›The agent had to visit /prestation to find the agent service pricing (1,990 € setup, 300 €/month), which was not disclosed on the homepage, showing asymmetric information architecture.
• ›Training pricing is entirely absent from the site; the agent correctly noted this gap rather than inferring or guessing prices, demonstrating sound handling of missing data.
• ›The site conflates training funding eligibility ('100% finançable') with no out-of-pocket cost on the homepage, but the /formations page clarifies this depends on profile and scheme—the agent resolved this contradiction by citing both claims and noting the conflict.
• ›The agent correctly sourced its final response to three specific pages (boosterlabs.fr, /prestation, /formations) and explicitly noted it did not use web search, demonstrating transparent attribution.
• ›Course content is described only by title, duration, and high-level topics; the site references a programme PDF that was not fetched or linked. The agent flagged this absence and did not substitute model knowledge, showing appropriate epistemic restraint.