✔ read-only environment diagnosis identifies actual runtime, missing dependencies and sealed inputs without installing or running scripts (7417.3386ms) ✔ diagnosis rejects unsupported managers, lock integrity and credential configuration without reading config content (7308.1121ms) ✔ approved real npm ci runs only in owned worktree, skips scripts, retains logs and binds later actual tests (14880.2298ms) ✔ explicitly allowing lifecycle scripts executes sealed root script and preserves the separate preparation result (11079.6146ms) ✔ stale lockfile, tampered plan/config, cross-session identities and mismatched download sources never dispatch npm (7779.604ms) ✔ registry failure keeps original npm exit/log and performs no automatic fallback or replay (10665.0492ms) ✔ stopping a hanging dependency download records only its owned process and preserves stopped logs (12454.18ms) ✔ restart retains unknown preparation, blocks writes and checks, and requires explicit read-only reconciliation (12354.2676ms) ✔ HTTP environment flow survives refresh, requires explicit approval and preserves task and artifact ownership (18715.7269ms) ✔ Python environment diagnosis reads declarations and runtime identity without executing project code or installing packages (797.8884ms) ✔ dependency snapshots reject external links and incomplete budgets instead of claiming environment equivalence (666.6091ms) ✖ model-requested environment preparation uses the original approval boundary and never treats preparation as passed engineering checks (12625.1077ms) ✔ timeout and offline cache miss retain actual failures and new attempts require newly sealed plans (15462.5984ms) ✔ evaluation denominator distinguishes attempted unsealed candidates from unexecuted samples (0.7405ms) ✔ evaluation summary retains sealed outcomes and stopped or failed original attempts (0.4211ms) ✔ evaluation contract fixes bounded samples, source paths, checks and aggregate budget (751.9058ms) ✔ same-source workspaces seal real commands before execution and preserve original staged and unstaged bytes (20323.131ms) ✔ actual original failures and candidate success are both preserved, then adoption requires parent review, checks and Git (85416.294ms) ✔ candidate cannot shrink tests and changing original test bytes rejects execution before any model request (28883.822ms) ✔ missing upstream usage remains unknown and does not become a free run or invented cost (39002.5187ms) ✔ missing raw check logs preserve unknown result and original sample denominator (44361.1075ms) ✔ source and model changes never silently replace an existing batch input (3535.9512ms) ✔ model revision change before preparing a fixed sample cannot silently use the replacement connection (3383.4208ms) ✔ HTTP evaluations require host token, original fingerprint and session ownership; worker API cannot bypass evaluation flow (22594.9298ms) ✔ sample queue cannot skip or reset attempted samples, and reports freeze further writes (25561.0335ms) ✔ stop cancels only current candidate and preserves partial results instead of replaying a request (23444.8096ms) ✔ restart recovers interrupted evaluations as unknown and keeps completed checks and plan identity (26745.1086ms) ℹ tests 27 ℹ suites 0 ℹ pass 26 ℹ fail 1 ℹ cancelled 0 ℹ skipped 0 ℹ todo 0 ℹ duration_ms 324266.4815 ✖ failing tests: test at tests\environment.test.mjs:120:1 ✖ model-requested environment preparation uses the original approval boundary and never treats preparation as passed engineering checks (12625.1077ms) TypeError [Error]: Cannot read properties of undefined (reading 'status') at TestContext. (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/tests/environment.test.mjs:132:90) at async Test.run (node:internal/test_runner/test:1409:7) at async Test.processPendingSubtests (node:internal/test_runner/test:974:7)