What problem does it solve? It keeps a personal applicant wiki about five Moscow universities accurate and hallucination-free by collecting budget/paid seats, passing scores, tuition, and military training center data only from official admission websites, with every number linked to a source. ## Core Features & Use Cases - Structured wiki generation: Creates and fills 13 markdown files (university cards, program cards, comparison tables) covering directions 09.03.01–09.03.04 and applied AI. - Anti-hallucination sourcing: Enforces a whitelist of official university domains, requires per-fact citations with access dates, and records "no data" instead of guessing. - Four-level fetch cascade: Falls back from WebFetch to KVM curl, headless Chrome, and Wayback Machine/PDF orders when sites use CAPTCHAs or React SPAs. - Use Case: Ask to fill the BMSTU card, then run compare to regenerate the master tables of passing scores and tuition across all five universities. ## Quick Start Ask the agent to run the highschool-research init command and then fill university bmstu to create the wiki skeleton and populate the first university card.
