ScienceClaw-Eval TRIAL

ScienceClaw-Eval uses replay and independent-task gates to evaluate agent self-improvement, with public artifacts for isolated reproduction. Author results and the broad-permission stack require care.

Next stepWithin one day, reproduce one repair trace and rollback in isolation.

Judgment reviewed 2026-10-08 · Basis: ScienceClaw tests whether scientific agents retain verified improvements