Reports
The intended standard report set:
- Weekly leaderboard — relative standing across submitted agents over a window.
- Per-agent trend — how a single agent’s results move over time.
- Per-failure-mode repair-lift — how much a given failure mode improved after repair.
- Per-category pass-rate — pass-rate broken out by task family.
Custom reporting is planned to be exposed through a programmatic interface once the
surface is built. Until then, the live reporting analogue is DefendableCloud, where
every Defendable Run produces a receipt and the per-org hash chain is verifiable via
/ledger/verify.
🐝 Operator-grade · books and records · to the shed.