Investigated 6 PRs (5 with full hydration, 1 preview). Upstream merged PRs (#6189, #5042) and focused fork PRs (#1, #4) are clean, well-tested parser/engine fixes with no network, credential, or CI changes in visible hunks. PR #4 adds is_saddled fields mirroring the existing is_suspected pattern across crates/engine/src/game/{ability_utils,derived_views,effects/mod,filter,game_object,quantity}.rs and includes adversarial test coverage. PR #1 extends RevealHand in crates/engine/src/game/effects/reveal_hand.rs with a controller-scoped opponent fallback and parser tags in oracle_effect/imperative.rs, backed by hostile test fixtures ensuring the controller's own hand is never inspected. PR #6189 introduces spell_object_id provenance and exhaustive integration tests in crates/engine/src/game/enters_with_unless_runtime_tests.rs for unless-condition replacement parsing. No malicious signals in any visible hunk. However, two fork-internal PRs (#8 with 312 files and #6 with 176 files) are massive branch-sync changes that include GitHub Actions workflows, deployment scripts, Tauri permission manifests, and dependency lockfiles. Patch evidence for those PRs was heavily truncated to fit the latency budget, so the highest-risk surfaces (CI YAML, deploy pipelines, native bridge code, lockfile diffs) could not be adversarially reviewed at the hunk level. Because the truncation gap falls directly on CI and supply-chain files, this shard cannot confidently clear the contributor. Investigator 2 shard: 5 PRs examined. Three upstream parser fixes (#6622, #6232, #6284) show clean, well-tested engine changes with no suspicious signals. Two PRs in the contributor's own fork (#5 and #7) are massive (193 and 61 files) and heavily patch-truncated. For #7, only CI workflow patches were visible; the core parser file (condition.rs) and 50+ other files were truncated. For #5, only two skill documentation patches were fully visible; the remaining ~190 files—including new native-engine modules (native_engine.rs, native_bridge.rs), Tauri permission expansions, Cargo/Node lockfiles, and multiple new CI workflows—were unseen. The visible CI hunks are benign structural refactors, but the unreviewed surface is too large to rule out malicious code. No cross-repo burst (all Phase project). No concrete backdoor evidence in any visible patch. Verdict is inconclusive because the highest-risk surfaces of the two largest PRs could not be inspected.