Per-card smoke tests for all 87 Silent cards (base + upgraded), split into 6 thematic buckets so parallel agents could work on them without stepping on each other: Bucket 1 - Basic Attacks (15 cards) Bucket 2 - Multi-hit & Shiv Attacks (15 cards) Bucket 3 - Conditional/Burst Attacks (12 cards) Bucket 4 - Defensive & Draw Skills (15 cards) Bucket 5 - Utility & Card Manipulation (16 cards) Bucket 6 - Powers & Poison Suite (14 cards) Each test runs an isolated combat against BigDummy, plays the card, and asserts on HP/block/energy/power deltas. Four reusable patterns in TestHelpers (SingleCardTest, PowerThenPlayTest, SequenceTest, PreloadHandTest) plus a Pass/Fail/Crash/Skipped result classification so harness bugs surface as Crash separately from "test got the wrong number". Run via: dotnet run -c Release -- silent-tests Current state: 166 pass, 0 fail, 0 crash, 8 skip. Skips are all genuine harness limitations (RandomEnemy targeting needs multi-enemy harness; MultiplayerOnly cards like Flanking need a RunManager shim; CombatRoom-gated cards like TheHunt need a real CombatRoom on the run state; CardSelectCmd discard needs hand fillers). Documented per skip. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>