Sixteen agent harnesses × 27 capabilities = 432 scored cells — every cell sourced to the vendor’s own documentation, with the full peer-review audit trail. Scores measure documented operational breadth (0–2 per capability): what a project documents and ships, not how well it codes.
Scores are not a ranking. A deliberately lean core scores lower on purpose, and a high total is not a verdict on quality — read the Method tab inside the map before quoting a number.
Open the map full size in a new tab · static exports: checklist (Markdown) · checklist (CSV, one row per cell) · peer-review audit trail