HANDBOOK.md: Results as Published

HANDBOOK.md tests whether an agent follows a long company policy document while doing multi-step tool work. Each of the 65 tasks sits in a self-contained company environment with a handbook of 20 to 124 pages and mock email, chat, calendar, issue-tracking and commerce services reached over the Model Context Protocol. A trial passes only when every programmatic criterion is met (824 criteria in total).

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.