✔ Messages builds native system/tool history, bounds images, and seals a distinct mixed-pool variant (6.9206ms) ✔ Messages URL and native parameter contracts preserve other protocol paths and hard output budget (0.5573ms) ✔ native usage includes both cache buckets once, retains raw and rejects invalid counters (0.1533ms) ✔ real HTTP Messages JSON executes one tool, retains opaque history, actual URL and cache accounting (3472.2818ms) ✔ real HTTP Messages SSE executes one tool, retains opaque history, actual URL and cache accounting (3214.277ms) ✔ Messages api-key authentication is explicit; mismatched model remains blocked with original usage (5.8812ms) ✔ Messages bearer authentication is explicit; mismatched model remains blocked with original usage (4.1845ms) ✔ Messages redirect is not followed, either credential stays at the original server (3.6778ms) ✔ strict native SSE reads the nested start model and retains mismatch source without retries (3.125ms) ✔ a closed native tool block cannot execute before message_stop, then executes exactly once (3437.2973ms) ✔ mixed Chat/Messages pool seals both variants and preserves one completed tool on failover (5728.1569ms) ✔ native JSON rejects incomplete stops, duplicate tools and hosted content (0.2702ms) ▶ strict SSE validates all blocks, terminal, event names, UTF8 and tool JSON before dispatch ✔ reject native stream 0 (1.2446ms) ✔ reject native stream 1 (0.2565ms) ✔ reject native stream 2 (0.1893ms) ✔ reject native stream 3 (0.2623ms) ✔ reject native stream 4 (0.1775ms) ✔ reject native stream 5 (0.1731ms) ✔ reject native stream 6 (0.1323ms) ✔ reject native stream 7 (0.125ms) ✔ reject native stream 8 (0.1771ms) ✔ strict SSE validates all blocks, terminal, event names, UTF8 and tool JSON before dispatch (3.6049ms) ✔ upstream SSE error retains original type/request ID and initial usage, never retries (5.0795ms) ✔ stop cancels the owned stream and retains partial consumption without running tool or replaying (63.1986ms) ✔ saved Messages import preserves auth choice, rejects endpoint duplication, and keeps source file untouched (796.6827ms) ✔ large source receipts use cursor-preserving invalidation without filling the event stream or changing the session (5.5041ms) ✔ small event updates retain the current session and actual cursor (0.135ms) ✔ the bundled local service enforces host, origin, token and JSON object contracts (2335.0186ms) ✔ an in-flight connection probe prevents configuration changes and another probe until its actual result returns (2500.194ms) ✔ API review applies verified content, detects conflicts, exports sessions and keeps keys in memory (4905.0536ms) ✔ a real bundle restart preserves proposals, interrupts old approvals and requests fresh consent on resume (12363.2312ms) ✔ runtime diagnostics and model-response validation use actual records and preserve the saved connection (2204.0479ms) ✔ persistent event replay is read-only and keeps run/artifact identity across reconnects (8508.7194ms) ✔ real browser HTTP plans, original field pages and PNG bytes preserve ownership and survive host restart (8411.3937ms) ✖ self harness browser tools require approval and feed verified actual page fields back to the original Task (16110.9832ms) ✔ browser host recovery retains unknown root outcome, blocks replay, and requires an explicit owned acknowledgement (3622.2194ms) ✔ late HTTP stop of a finished browser job cannot interrupt a later job in the same Task (7960.7089ms) ✔ actual failed page blocks source-passed delivery; a new passing page permits a bound commit and preserves both original records (20875.835ms) ✔ a build-only change after Git review invalidates the actual browser evidence and leaves HEAD and the protected index intact (8878.7864ms) ✔ actual DOM controls include validated unique CSS selectors across repeated parent buttons and escaped labels (7374.818ms) ✔ browser field coverage distinguishes sections and Run windows while retaining original source identity (0.4994ms) ✔ repeated original browser pages pause a live Task without re-opening the browser or breaking tool pairing (5394.1247ms) ✔ browser plan rejects outside builds, invalid interaction scope and unbounded steps before execution (0.9867ms) ✔ real owned Edge interactions, assertions, raw PNG and stopped process survive exact result reuse and host reopen (4103.2666ms) ✔ failed real page assertion preserves actual value and forbids silent retry of the original plan (8540.5642ms) ✔ build/source mutation and sealed step tampering reject before browser dispatch; earlier task cannot execute plan (1689.2822ms) ✔ outside network never reaches another local service, and original JavaScript/console failures remain failures (3227.7997ms) ✔ stopping a real in-flight browser preserves finished steps, closes its owned root and never replays (3100.5875ms) ✔ 原PNG损坏使原结果校验失败,不能以已有通过状态复用 (3116.5306ms) ✔ fixed browsing contracts reject local host APIs, credential URLs, protocols, cross-origin and arbitrary execution (0.9995ms) ✔ attested desktop queue preserves same-tab observations, original PNG bytes and approval-bound one-shot actions (1618.6872ms) ✔ page data cannot grant domains, interaction or other Task scope; nonce and ownership remain fixed (1310.7955ms) ✔ unknown dispatched submission blocks even after a new observation, rejects late success and retains explicit acknowledgement (1524.3801ms) ✔ downloads require exact approved address, remain original binary artifacts, and cannot reuse consumed observation (1093.103ms) ✔ revoking shared persistent profile closes all its tabs and clears only that profile after original desktop acknowledgement (2156.6103ms) ✔ service recovery never reopens old tabs or replays dispatched actions and keeps unknown original receipts (1029.5664ms) ✔ agent browser flow waits for original Task/Run approval, dispatches once, and returns paired original results (5135.2371ms) ✔ changed approved input and stopped dispatched agent action cannot be accepted or replayed (1609.6903ms) ✔ browser inventory gives bounded metadata and original references instead of repeating full website histories (1235.6559ms) ✔ receipt storage failure settles the original dispatched action as unknown and never hangs or accepts late success (1701.5211ms) ✔ input that fits the engine window but exceeds remaining credit diagnoses the credit limit (0.501ms) ✔ window overflow and historical pauses retain distinct and honest diagnoses (0.0802ms) ✔ R observed reservation shortage keeps the window-fit diagnosis separate from Run/Task credit (0.0979ms) ✔ a budget pause persists concrete text/tool/history/image breakdown and never dispatches an oversized request (3178.2026ms) ✔ explicit frozen preview resumes the same Task in a new Run, preserves completed tools and does not change global config or selected model (4173.7295ms) ✔ insufficient budgets, silently clamped capacity and arbitrary config fields reject without a physical request or mutating task budget (2705.2227ms) ✔ changed session, changed registry, expiry and altered frozen preview all require a new explicit preview (3609.1524ms) ✔ preview preserves unknown invocation status and increasing Task allowance is committed only with confirmed new Run (4886.4996ms) ✔ batch APIs enforce ownership and review fingerprints, apply once, and persist the completed batch across restart (6634.8177ms) ✔ delivery reports bind actual batch identity and reject corrupted snapshot evidence without changing project files (4322.1631ms) ✔ a compatible model can seal proposals through the real HTTP loop while application stays behind user review (8267.2292ms) ✔ a sealed batch at the step limit hands off and continues the same Task into actual checks without replaying saved tools (19200.452ms) ✔ applied batches cannot turn cancellation, other budgets, restart or unknown commands into a review continuation (2498.3202ms) ▶ complete sealed batch hands off after all tools settle with no next model request ✔ complete (4985.1445ms) ✔ later-error (5936.6898ms) ✔ complete sealed batch hands off after all tools settle with no next model request (10922.7275ms) ▶ partial, changed, failed or cancelled batches cannot claim complete review handoff ✔ partial (5727.9139ms) ✔ new-proposal (5088.9095ms) ✔ failed (5241.4825ms) ✔ conflict (5388.9402ms) ✔ cancelled (4893.4765ms) ✔ unknown (5514.1401ms) ✔ tampered (5340.639ms) ✔ partial, changed, failed or cancelled batches cannot claim complete review handoff (37196.8257ms) ✔ actual process exit at file-batch prepared preserves durable identity and never replays writes (2152.1565ms) ✔ actual process exit at file-batch application_started preserves durable identity and never replays writes (2129.2665ms) ✔ actual process exit at file-batch write_intent preserves durable identity and never replays writes (2403.1453ms) ✔ actual process exit at file-batch staged preserves durable identity and never replays writes (2102.0432ms) ✔ actual process exit at file-batch replaced preserves durable identity and never replays writes (2749.511ms) ✔ actual process exit at file-batch file_recorded preserves durable identity and never replays writes (2634.7802ms) ✔ actual process exit at file-batch finished preserves durable identity and never replays writes (2638.183ms) ✔ actual process exit at file-batch rollback_started preserves durable identity and never replays writes (3581.3904ms) ✔ actual process exit at file-batch restore_intent preserves durable identity and never replays writes (3484.057ms) ✔ actual process exit at file-batch restore_staged preserves durable identity and never replays writes (3507.0916ms) ✔ actual process exit at file-batch restored preserves durable identity and never replays writes (3230.0401ms) ✔ actual process exit at file-batch restore_recorded preserves durable identity and never replays writes (3380.0581ms) ✔ actual process exit at file-batch rollback_finished preserves durable identity and never replays writes (3403.8842ms) warning: in the working copy of 'README.md', LF will be replaced by CRLF the next time Git touches it warning: in the working copy of 'src/two.mjs', LF will be replaced by CRLF the next time Git touches it ✔ sealed batches preserve bytes until review, then apply and explicitly restore BOM/CRLF/new files (2788.8717ms) Switched to a new branch 'owner-branch' ✔ a conflict in the last file or Git branch prevents writing every file in the batch (3273.2125ms) ✔ duplicate canonical paths, ancestor collisions, links, dependencies and Windows aliases never form a batch (1410.1749ms) ✔ fingerprints, changed proposals and corrupt snapshot artifacts cannot write or restore any target (2409.6815ms) ✔ a failure after replacement restores only batch bytes and never retries an old application (2277.7817ms) ✔ rollback preserves owner edits, blocks other project writers, and explicit acknowledgement releases recovery (2229.6237ms) ✔ an unattempted owner-created file identical to a suggestion remains external and is never removed (1717.5504ms) ✔ active batch requests are serialized and cancellation leaves a verifiable ended batch (1621.859ms) ✔ batch application and rollback preserve unrelated Git index and worktree bytes (4320.6866ms) ✔ recovery reuses exact staged bytes but retains a conflicting temporary file for owner inspection (3646.1592ms) ✔ one Task can seal pending proposals from earlier Runs while snapshots retain each source Run (3406.8928ms) ✔ engineering inspection preserves a dirty Git baseline and treats project rules as data (2702.6266ms) ✔ real npm tests and build produce counted results, log artifacts and persistent verification (3799.2022ms) ✔ an actual failing test retains the nonzero exit code and failed count (2860.0872ms) ✔ unchanged automatic checks return original failure evidence, while explicit user reruns and changed source still execute (6224.3144ms) ✔ agent loops cannot hide identical failing checks behind new plan IDs or unrelated reads (10315.4511ms) ✔ zero tests, skipped tests and unrecognized stdout cannot become verification passes (8939.5065ms) ✔ a successful build with no selected tests does not verify the task (2235.8467ms) ✔ stale check plans and forged script names execute no project commands (1954.9041ms) ✔ applying a later proposal invalidates the prior passed engineering snapshot (3018.8302ms) ✔ stopping an actual check terminates its child process and blocks concurrent jobs (2428.3681ms) ✔ check timeout records a failure with its actual termination result (3279.4059ms) ✔ source edits during a successful check make the result stale (3487.4473ms) ✔ bounded large logs retain the test footer and readable artifact instead of inventing counts (2202.7251ms) ✔ model-selected checks require a run-bound decision and denial creates no check job (4315.1325ms) ✔ check APIs expose owned plans and jobs and preserve the authoritative results across restart (6789.2904ms) ✔ reviewed changes continue the original task without reopening a generic completed run (3547.9944ms) ✔ concurrent review-continuation API requests accept one run and retain the goal after restart (9638.0532ms) ✔ the model reads owned failure logs without running the check again (3206.7473ms) ✔ process exit at check queued preserves results and never replays commands (1414.769ms) ✔ process exit at check result_committed preserves results and never replays commands (2185.9866ms) ✔ model code tools keep Task/Run ownership, index evidence and exact proposals across storage recovery (6246.2446ms) ✔ bundled code APIs isolate project results, validate queries and create review-only patches with conflicts (2846.961ms) ✔ code tool arguments reject malformed ownership-independent inputs before side effects (0.2936ms) ✔ observe-only device authority cannot read, configure or clear project parsing caches (1169.6086ms) ✔ pairing is short lived, single-use, generation-bound and validates exact permissions (838.1304ms) ✔ device transaction failure neither consumes the ticket nor leaves a credential (783.121ms) ✔ devices have project and action boundaries; rotation, expiration, revocation and host changes reject old authority (1126.5411ms) ✔ persistent registry stores hashes and a bounded audit, never raw credentials (797.4511ms) ✔ real TLS verifies IP SAN, CA chain, encrypted restart and public fingerprint (946.1448ms) ✔ TLS sealing failure leaves listener closed and no plaintext or successful TLS registration (853.1165ms) ✔ HTTP pairing is concurrent single-use; secure cookie and strict origin/host checks (874.8764ms) ✔ gateway never exposes management routes, arbitrary assets or unpaired project data (1080.4938ms) ✔ authorized project events paginate from original seq without leaking other project events or payloads (1468.0997ms) ✔ remote task projection excludes raw model history, host configuration, tool input and credentials (1197.0186ms) ✔ permission checks apply to direct remote mutation, not only UI controls (844.1345ms) ✔ approval fingerprint and exact task/revision are bound; duplicate requests never execute twice (1433.4396ms) ✔ unknown original operation is recorded and never replayed after response failure (1135.5444ms) ✔ revocation and gateway closure reject old access without invoking runtime stop (898.7077ms) ✔ rate limits reject pairing brute force and do not persist candidate secrets (774.8743ms) ✔ real Windows CurrentUser DPAPI persists only a sealed CA and reopens it (1926.3946ms) ✔ mobile approval executes the original Runtime command exactly once and rejects the later desktop decision (4442.2339ms) ✔ desktop rejection wins a concurrent mobile request and original command remains unexecuted (2502.7943ms) ✔ phone stop waits for the original process exit and prevents later heartbeats and command replay (3132.2302ms) ✔ Run replacement during parent-stop wait is rejected rather than stopping the new Run (884.8938ms) ✔ diff pages preserve source SHA; report reads verify blobs and exclude unrelated artifact kinds (1011.9823ms) ✔ hot upgrade refuses an active gateway and restores closed gateway ownership on failure (863.0151ms) ✔ damaged device permissions or expiry fail closed and never overwrite the original registry (845.9167ms) ✔ device revocation destroys its actual authenticated TLS connection (1032.5002ms) ✔ companion observes only current Task screenshots, verifies original PNG bytes and loses access immediately on revoke (1355.1641ms) ✔ mobile browser observation is current Task/project scoped, preserves original PNG and exposes no browser control or profile data (1129.8233ms) ✔ actual config HTTP failure preserves the committed selection, config and volatile credential before a successful retry (1028.0856ms) ✔ controlled installation pins the native kernel and bridge independently of the existing connection-probe identity (1014.6676ms) ✔ controlled profiles disable every native project/terminal provider and native model/retry route using the actual include patch contract (0.6893ms) ✔ owned tool catalogues seal full schemas canonically and reject duplicate or malformed declarations (4.9264ms) ✔ bridge authentication, current Run, process, SDK session and browser-origin checks reject before dispatch (691.9354ms) ✔ bridge duplicates, unknown tools, malformed arguments and source changes are refused before a second side effect (592.0143ms) ✔ bridge model requests bind the exact catalogue, route and output bound and preserve raw request/result identity (761.199ms) ✔ stopped and oversized bridge requests never reach a local model or tool callback (885.5092ms) ✔ a Windows virtual data-directory name binds new storage and credentials to its actual physical root (2131.9437ms) ✔ physical data-root adoption still rejects explicit junctions at the directory or any ancestor (760.9312ms) ✔ actual Windows CurrentUser DPAPI roundtrips owned UTF-8 and rejects damaged bytes and wrong entropy (1266.5053ms) ✔ only explicitly supplied keys persist; new provider restores original revisions and public views contain no key (2659.6624ms) ✔ registry and cipher commit failure restores original routing, keys, block and every database row (1840.6193ms) ✔ a two-key import with a protection failure rolls back all registry and cipher entries and retains its preview (1949.527ms) ✔ connection revision changes retain the original saved block and report a source mismatch on restart (2327.5298ms) ✔ disable skips real restart loading, explicit enable restores, delete preserves memory and forget removes both (2552.4832ms) ✔ corrupt saved block and another SID do not load; failures are visible without returning ciphertext or plaintext (2600.6409ms) ✔ legacy persistence requires a new explicit key and binds config and credential revisions; stale requests cannot save (2471.5392ms) ✔ owned plaintext keys never reach SQLite pages, WAL, config or public execution exports (1426.8639ms) ✔ reviewed application invalidates old checks, names current SHA and advances through a new approved check to Git preview (6667.1834ms) ✔ skip-tests, foreign task, unknown operations, diagnostic and parent review retain their boundaries (2879.9307ms) ✔ before the first change, real baseline failures persist across later reads; passed baselines and changed sources cannot certify delivery (7294.7437ms) ✔ actual process exit at Git queued never replays a commit and can reconcile the durable result (10779.9552ms) ✔ actual process exit at Git preparing never replays a commit and can reconcile the durable result (11769.8099ms) ✔ actual process exit at Git index_locked never replays a commit and can reconcile the durable result (11787.2765ms) ✔ actual process exit at Git prepared never replays a commit and can reconcile the durable result (12438.4861ms) ✔ actual process exit at Git ref_prepared never replays a commit and can reconcile the durable result (13478.1339ms) ✔ actual process exit at Git publishing never replays a commit and can reconcile the durable result (13149.9454ms) ✔ actual process exit at Git committed never replays a commit and can reconcile the durable result (11073.9433ms) ✔ actual process exit at Git index_published never replays a commit and can reconcile the durable result (9723.3983ms) ✔ actual process exit at Git finished never replays a commit and can reconcile the durable result (9353.3296ms) ✔ an unknown Git result cannot overwrite an externally changed index or delete an unrelated lock (9280.1273ms) ✔ known publication with a later source edit retains the commit and marks the delivery for inspection (9358.3152ms) ✔ CRLF normalization, multiple applied revisions and unrelated special index flags preserve exact bytes and ownership (10851.6884ms) ✔ adding new durable record kinds to an existing hot Store preserves all existing Maps and records (8538.6954ms) ✔ an unrelated owner deletion is fingerprinted and preserved while linked task paths are rejected (13953.4565ms) ✔ actual failed tests become a reviewed local commit and evidence report while preserving unrelated index and files (12509.7187ms) ✔ selected preexisting unstaged, staged and untracked owner edits cannot enter a task commit (14166.5822ms) Switched to a new branch 'owner-branch' ✔ source, unrelated stage and branch changes after review invalidate a commit plan without changing HEAD (18716.072ms) Switched to a new branch 'delivery-branch' ✔ a later Git metadata change keeps matching source evidence valid but a real source edit requires checks (9449.4475ms) ✔ a corrupt passed-check log and a pending proposal block Git delivery instead of inventing evidence (5166.8722ms) HEAD is now at c01f5cc Sealed delivery fixture baseline ✔ Git filters, merge state, detached HEAD and old task ownership are explicitly rejected (21439.8249ms) ✔ local snapshot commits disable repository hooks and signing and isolate the Git child environment (8533.1882ms) ✔ stopping before publication cancels safely; stopping after publication preserves the actual completed commit (15606.6955ms) ✔ delivery API confirms exactly one owned plan, serves verified reports and preserves results across restart (12199.4924ms) ✔ the model can prepare an owned Git review and evidence report while publishing remains a separate user action (10343.1033ms) ✔ HTTP Git review pairs deferred calls, survives reload and never requests more work until the saved preview is resolved (10504.4469ms) ✔ a reviewed Git preview can be revised or withdrawn without publishing or losing the current Task (8726.3573ms) ✔ actual Electron debug argument ordering retains exactly the compiled entry and rejects extra or repeated actions (1.0978ms) ✔ native bridge refuses other origins, subframes, other windows and unsupported inputs before connecting (535.6535ms) ✔ desktop preview opening requires a currently registered project preview and rejects iframe callers (693.4009ms) ✔ file reveal binds the current project, task, path, UTF-8 bytes and original SHA without executing the file (626.9406ms) ✔ file changed during asynchronous task recheck is not revealed as the old bytes (675.2725ms) ✔ desktop browser profile and aliases stay outside project inventory and file access (651.6063ms) ✔ native picker cancel and origin change after dialog publish no project registration (559.7233ms) ✔ pending native action is not dispatched twice, and current task review returns its original seq (640.5021ms) ✔ native stop cancellation and changed original host never stop a replacement process (650.2883ms) ✔ window register rejects host PID, stale installation, extra fields and missing creation identity (0.3486ms) ✔ a rebuilt source archive binds actual generated client assets and retains its original running identity after disk changes (1938.5009ms) ✔ host challenge proof rejects a stale challenge, changed identity and changed signature (0.8535ms) ✔ stop confirmation binds the original instance, creation and loaded runtime (0.1664ms) ✔ state paths reject escaping, linked ancestors, ambiguous hardlinks and damaged JSON (968.6488ms) ✔ failed registration publishes no initialized memory and can complete after its actual inputs are supplied (2408.6094ms) ✔ host shutdown cancels an un-signalled HTTP model request and refuses later requests until recovered (1.3691ms) ✔ host shutdown cancels model-pool acquisition and releases queue ownership without inventing attempts (600.1567ms) ✔ full candidate validation returns the exact loaded-input host identity and rejects undeclared or altered files (1049.918ms) ✔ fast per-file scan still rejects a junctioned subdirectory, a hard-linked file and a stream path with the original bytes intact (962.917ms) ✔ installation requires separate absolute program/data directories and a candidate in its own versions namespace (1198.7429ms) ✔ inventory rejects a new hardlink on an optional file after its full-manifest scan and accepts the untouched candidate after removal (948.1029ms) ✔ staging retains the original data and pointer without querying or restarting a host (3014.361ms) ✔ a post-pointer failure restores every original pointer byte and persistent row; a first-install failure removes only its published pointer (10819.9612ms) ✔ changed active manifest and concurrent pointer changes are rejected without overwriting the unknown record (4693.9663ms) ✔ an unregistered non-HTTP listener blocks preflight even when service-info is absent (2825.5298ms) ✔ uninstall archives its original pointer and preserves every data row/artifact instead of recursively deleting data (6425.378ms) ✔ preflight rejects an unrelated program root or changed native bootstrap while preserving its original bytes (4555.6187ms) ✔ a changed owned bootstrap follows its candidate and a post-pointer fault restores both original bytes (16800.2755ms) ✔ an external bootstrap change after preflight is preserved and never overwritten as an update or rollback (8210.8844ms) ✔ owned startup retries only read-only observation after a slow health response and attests the same original child (3815.0788ms) ✔ unverified startup expires with its original live child and never treats timeout or an unrelated health PID as proof (5713.5198ms) ✔ diagnostic failures and repeated questions retain the original execution error, task, checks and read-only authority (3008.1439ms) ✔ diagnostic binding survives reopen, rejects foreign Task references and detects original error or artifact tampering (966.5661ms) ✔ a new Task never adopts a previous Task error from session.lastFailure (617.0004ms) ✔ verified research and explicit no-test scopes do not demand an unrelated test job or claim engineering success (3144.0529ms) ✔ unknown scope and unknown operations remain actionable even after a diagnostic answer or read proof (1207.8403ms) ✔ delivery status HTTP is read-only, selection-bound and equal to the saved report after host restart (4523.3144ms) ✔ routine engineering facts remove duplicate success excerpts while complete diagnostics and errors remain available (715.038ms) ✔ a question about the previous workbench error receives local error context and current change states (2410.1536ms) ✔ an unknown model request stops, and explicit resume never replays the approved command (3610.4936ms) ✔ exhausted refused connections preserve tool results, and explicit resume adds no new user turn or replay (3305.6044ms) ✔ stopping during model backoff cancels further requests and clears the running request state (1693.4837ms) ✔ HTTP authentication errors are classified, redacted and never automatically retried (17.2198ms) ✔ temporary HTTP service errors retry and recover with recorded attempt counts (13.4289ms) ✔ invalid and oversized protocol responses fail once with a useful response-format diagnosis (8.2603ms) ✔ HTTP rate limits expose their source and obey the service retry interval before recovery (1007.3791ms) ✔ Windows commands preserve Chinese through a PowerShell pipeline into Node (849.0411ms) ✔ Windows shell syntax errors stay readable and native command exit codes are retained (1183.4222ms) ✔ review and completed states do not turn a generic continue click into a new model request (815.8697ms) ✔ old session data receives a truthful historical connection diagnosis without rewriting history (759.1014ms) ✔ engine HTTP checks bind the detected installation and current task, page original bytes and preserve them across host restart (5332.7614ms) ✔ HTTP stop seals a pending SDK call, rejects automatic replay and keeps failures attached to the original session (4281.088ms) ✔ sealed r1-off failure fits its original 16K config with authentic latest status and readable source (13.0453ms) ✔ sealed r1-on failure fits its original 16K config with authentic latest status and readable source (10.0288ms) ✔ sealed r2-on failure fits its original 16K config with authentic latest status and readable source (7.6635ms) ✔ sealed r2-off failure fits its original 16K config with authentic latest status and readable source (7.3522ms) ✔ archived native call pairs require exact saved arguments/results rather than model-written summaries (8.219ms) ✔ full system/user requirements and mandatory current failure can refuse rather than be trimmed (12.0476ms) ✔ small native input is byte-preserving; developer instructions remain mandatory after compaction (0.18ms) ✔ older failing checks and pending proposals remain visible after a later large file read (14.2207ms) ✔ bounded result keeps every check status and cuts Unicode only on whole characters (0.475ms) ✔ controlled engine HTTP freezes one child, pages original owned input/report, rejects foreign reads and prevents changed-source reuse (15198.217ms) ✔ a sealed whole-batch review pauses the owned native session without certifying engine completion or losing the saved host result (14577.6631ms) ✔ a changed review seal still closes the owned native process and cannot claim a verified review handoff (14618.8289ms) ✔ native text and call IDs preserve original result pairing while unsupported content fails closed (0.4212ms) ✔ one engine child freezes real installation identity before creating tasks and rejects a changed source (9017.8826ms) ✔ owned native tools return saved results, use real approvals/checks and account each actual request once across new engine Runs (28777.6963ms) ✔ forged native scope cannot change a parent file and a final completion claim cannot replace worker proof (14656.5356ms) ✔ duplicate native IDs are rejected without replaying a saved tool (15053.0091ms) ✔ native input exceeding the context budget is retained and sends no model request (10897.0092ms) ✔ bounded native sends verify original tool provenance and reject forged views or missing versions (14128.1501ms) ✔ stop retains a late reported model usage once and executes no tool from that output (11362.3683ms) ✔ a source change after the original model intent stops before native tools execute (11199.5409ms) ✔ an engine source change while approval is pending invalidates the old permission before any check starts (18136.7643ms) ✔ missing or noncompleted turn end cannot be adopted even with an owned receipt and assistant text (0.5895ms) ✔ closed root with an original live descendant remains unknown and retains its process tree (0.2926ms) ✔ host recovery reserves original pending attempts exactly once and explicit review preserves unknown outcomes (14765.0196ms) ✔ corrupted model reservation cannot reduce recovered budget or permit a new engine request (13977.1613ms) ✔ engine detection seals the actual CLI, SDK, profile and Node and rejects mixed versions without launch (755.819ms) ✔ owned engine probe waits for real ready identity, verifies six native responses and saves immutable task evidence (1478.0644ms) ✔ an invalid startup nonce is rejected before initialize and preserves the process close and partial report (1796.9676ms) ✔ SDK identity and session directory mismatches retain raw responses and never become verified (4479.0304ms) ✔ cancel during a pending initialize closes only the owned process without sending an unsupported MCP cancellation (1656.2749ms) ✔ cancel before readiness sends no protocol request and a user retry creates a different process and record (1954.9497ms) ✔ source or scratch changes and unexpected model traffic fail after preserving the actual observed result (4595.9465ms) ✔ unowned idle notifications stay quarantined and server requests receive a rejection without granting operations (2818.3392ms) ✔ engine probe environment excludes parent secrets and NODE_OPTIONS while preserving operating system paths (1800.1727ms) ✔ foreign sessions and altered reports or task bindings cannot reuse verified engine evidence (1539.6377ms) ✔ restart never replays an active engine check or stops a historical PID, and keeps prior task state (1417.6968ms) ✔ model argument validation and runtime activity do not grant external engine prompt or project execution (673.1155ms) ✔ a follow-up question receives the actual failed engine check from its original Task and keeps its report unchanged (2280.1716ms) ✔ unfixed dependency URLs and invalid integrity are reported without exposing URL credentials or allowing a plan (873.7985ms) ✔ hot upgrade refuses an active dependency operation or evaluation before changing runtime objects (0.5256ms) ✔ read-only environment diagnosis identifies actual runtime, missing dependencies and sealed inputs without installing or running scripts (6485.2796ms) ✔ diagnosis rejects unsupported managers, lock integrity and credential configuration without reading config content (5332.9846ms) ✔ approved real npm ci runs only in owned worktree, skips scripts, retains logs and binds later actual tests (10696.8064ms) ✔ explicitly allowing lifecycle scripts executes sealed root script and preserves the separate preparation result (8533.1088ms) ✔ stale lockfile, tampered plan/config, cross-session identities and mismatched download sources never dispatch npm (7077.9823ms) ✔ registry failure keeps original npm exit/log and performs no automatic fallback or replay (8422.7342ms) ✔ stopping a hanging dependency download records only its owned process and preserves stopped logs (9430.6853ms) ✔ restart retains unknown preparation, blocks writes and checks, and requires explicit read-only reconciliation (7261.3613ms) ✔ HTTP environment flow survives refresh, requires explicit approval and preserves task and artifact ownership (13202.8257ms) ✔ Python environment diagnosis reads declarations and runtime identity without executing project code or installing packages (883.7501ms) ✔ empty build cache directories keep dependency identity while added executable files invalidate it (1008.7462ms) ✔ dependency snapshots reject external links and incomplete budgets instead of claiming environment equivalence (722.7873ms) ✔ model-requested environment preparation uses the original approval boundary and never treats preparation as passed engineering checks (12503.7383ms) ✔ timeout and offline cache miss retain actual failures and new attempts require newly sealed plans (12747.9273ms) ✔ both evaluation worktrees can approve fixed dependency preparation and different actual dependencies reject candidate execution (32756.4645ms) ✔ evaluation denominator distinguishes attempted unsealed candidates from unexecuted samples (0.3259ms) ✔ evaluation summary retains sealed outcomes and stopped or failed original attempts (0.4437ms) ✔ evaluation contract fixes bounded samples, source paths, checks and aggregate budget (597.6684ms) ✔ same-source workspaces seal real commands before execution and preserve original staged and unstaged bytes (16526.6775ms) ✔ actual original failures and candidate success are both preserved, then adoption requires parent review, checks and Git (60625.0555ms) ✔ candidate cannot shrink tests and changing original test bytes rejects execution before any model request (21514.3615ms) ✔ missing upstream usage remains unknown and does not become a free run or invented cost (37863.3101ms) ✔ missing raw check logs preserve unknown result and original sample denominator (38744.5009ms) ✔ source and model changes never silently replace an existing batch input (3475.1232ms) ✔ model revision change before preparing a fixed sample cannot silently use the replacement connection (3313.9687ms) ✔ HTTP evaluations require host token, original fingerprint and session ownership; worker API cannot bypass evaluation flow (17908.4156ms) ✔ sample queue cannot skip or reset attempted samples, and reports freeze further writes (30960.4924ms) ✔ stop cancels only current candidate and preserves partial results instead of replaying a request (23971.4018ms) ✔ restart recovers interrupted evaluations as unknown and keeps completed checks and plan identity (22342.1934ms) ✔ JSON migration survives process exit before commit and retains exact legacy backups (3239.1639ms) ✔ process exit after_state rolls back state, event and outbox together (2007.332ms) ✔ process exit after_event rolls back state, event and outbox together (1874.5467ms) ✔ crash recovery at accepted retains execution identity and never replays side effects (1567.5591ms) ✔ crash recovery at queued retains execution identity and never replays side effects (2152.3426ms) ✔ crash recovery at approval_wait retains execution identity and never replays side effects (2945.9365ms) ✔ crash recovery at approved retains execution identity and never replays side effects (2813.9811ms) ✔ crash recovery at effect_unrecorded retains execution identity and never replays side effects (3366.0761ms) ✔ crash recovery at result_committed retains execution identity and never replays side effects (4714.5779ms) ✔ crash recovery at before_next_model retains execution identity and never replays side effects (5241.1726ms) ✔ state and event failures publish no success and preserve the last committed memory snapshot (1031.0308ms) ✔ tool result artifacts and event cursors retain project ownership across restart (2884.324ms) ✔ token budgets pause before tools execute and continuation keeps original task and accumulated usage (2330.5829ms) ✔ a time budget interrupts an unresponsive model and persists a resumable run (794.5335ms) ✔ a corrupted large result artifact cannot become a recovered success (1322.7812ms) ✔ hard-link aliases cannot expose protected bytes through reads, proposals, inventory or automatic project rules (1017.9224ms) ✔ ambiguous project hard links are explicit and ordinary independent file copies remain readable (665.5487ms) ✔ current-file HTTP receipts bind Task/Run and bytes, persist on restart and end without altering original runs or user Git data (5256.7283ms) ✔ file targets exclude referenced inputs and prohibited subsequent modifications (1.5596ms) ✔ full actual read and existing receipt complete a no-test file task without fabricating changes, checks or commits (2581.2213ms) ✔ partial pages, partial lines, wrong hashes and failed reads cannot be adopted; contiguous full paging can (1289.422ms) ✔ user edits and new requirements invalidate receipts, and a second task must explicitly adopt a source in its own project/session (1057.7948ms) ✔ adopting an unchanged file never satisfies an explicit modification or a new modification instruction (918.1169ms) ✔ required checks and stale passing flags cannot be bypassed by current file receipts (1055.8465ms) ✔ results plan proofs bind current file receipts while changes proofs still demand real applied change sets (817.1339ms) ✔ duplicate receipts cannot create progress indefinitely or reset no-progress reads (3670.0635ms) ✔ manual completion preserves the failed original run and refuses unknown operations (836.0311ms) ✔ completion refuses active work, files changed during inspection and a newly selected task (720.9076ms) ✔ unrelated files cannot stand in for named output paths, nor can a late apply satisfy newer instructions with an old proposal (1011.8858ms) ✔ delivery and mobile projections distinguish adopted bytes, new changes, no tests and no commit; mutations cannot rewrite receipts (1341.1956ms) ✔ proposals preserve project bytes until review; apply backs up the exact original bytes (609.1657ms) ✔ new nested files are created only on apply and unchanged content creates no proposal (472.4608ms) ✔ stale reads and changed baselines are rejected without overwriting external edits (882.1597ms) ✔ file paths reject traversal, absolute paths, credentials and outside junctions (696.8982ms) ✔ read and search expose bounded text scopes and reject non UTF-8 / oversized files (746.1416ms) ✔ project-local junctions cannot expose or create protected Git and credential files (607.5555ms) ✔ Windows trailing-dot and trailing-space path aliases cannot bypass protected paths (697.5181ms) ✔ a starter is copied into user data once; restarting from another program version preserves existing projects and editable bytes (1489.0337ms) ✔ startup does not insert a demo or rewrite/reorder an existing user project registry (572.1805ms) ✔ PNG bytes are fully decoded, metadata-free and bounded; JPEG EXIF orientation is applied to the sent PNG (104.4037ms) ✔ forged type, short headers, CRC damage, incomplete chunks, trailing payloads, animation and dimension bombs reject (23.8002ms) ✔ draft image ownership, withdrawal, reload and atomic Task binding retain original SHA and original bytes (983.4029ms) ✔ a failed session transaction restores draft status and cannot leave a Run consuming the attachment (604.657ms) ✔ draft limits, duplicate selection, source changes and cross-Task association reject before dispatch (1149.032ms) ✔ demo and legacy paths refuse image dispatch without creating a Task or consuming the draft (683.515ms) ✔ chat-completions sends one owned image per actual request, pairs original tools and verifies exact sealed bytes (2637.0188ms) ✔ responses sends one owned image per actual request, pairs original tools and verifies exact sealed bytes (2796.9747ms) ✔ mixed pool excludes unknown vision before lease and sends zero HTTP to the incompatible member (1841.1117ms) ✔ vision probe requires correct randomized pixel order and evidence expires when the connection revision changes (913.9711ms) ✔ HTTP upload/read/withdraw uses session authorization, returns byte SHA and refuses unknown ownership (724.9261ms) ✔ image accounting excludes Base64 as text but seals every original request byte and reserves image input separately (1.9315ms) ✔ actual model comes from owned consumption, excludes other Tasks and deduplicates the current Run (1.5452ms) ✔ mixed routes retain each original model/protocol/outcome and unknown use is never reported as zero (0.1374ms) ✔ zero dispatch, missing ledger, missing model and truncated Run history never borrow configured model identity (0.2165ms) ✔ repeated attempt identities across Runs and contradictory usage stay explicit instead of inflating successes (0.1195ms) ✔ observed route never certifies returned identity and retains mismatches and original source (0.1419ms) ✔ directory picker reads one level, preserves empty folders, skips credential and linked folders (580.6724ms) ✔ HTML preview serves encoded names and relative modules on an isolated loopback origin; closing revokes it (574.3394ms) ✔ HTML preview refuses credentials, hidden inventories, traversal, external junctions, hardlinks and unsupported file types (675.6249ms) ✔ runtime shutdown closes local preview ports and pending starts cannot escape shutdown (531.2016ms) ✔ idle file previews expire and release their loopback servers (734.4028ms) ✔ directory and preview HTTP controls require workbench authentication; project pages cannot read bootstrap (2415.4892ms) ✔ MCP registration stays disabled until an explicit real stdio probe and schema-based tool selection, with actual process cleanup (1887.9485ms) ✔ MCP probe holds the project operation lock until its owned process and report are settled (1406.0223ms) ✔ MCP validates the declared JSON schema subset and rejects unsupported schemas before external dispatch (0.9386ms) ✔ actual model MCP call approves long schema-valid arguments, validates the sent catalog and stores one original result without leaking environment secrets (4445.0181ms) ✔ denying an MCP call never starts another process or executes the external tool (3517.742ms) ✔ actual handshake drift rejects tools/call even if program entry hash is unchanged (4404.9694ms) ✔ cancelled dispatched external side effect stays unknown, closes its owned process and cannot replay on reopen (4128.4538ms) ✔ changed executable input, foreign project IDs and corrupted original contracts block enable and dispatch (1951.0209ms) ✔ registered MCP HTTP API uses fixed execution contracts and project-owned probe records with no model requests (3778.0328ms) ✔ unknown MCP pauses later tool calls and all project writers; fingerprinted acknowledgement clears only the review gates and explicit resume does not replay (5831.7745ms) ✔ MCP acknowledgement persistence failure restores both job and tool review gates; corrupt evidence cannot be acknowledged (4660.6557ms) ✔ actual host exit after an external effect restores unknown jobs without a fabricated report, allows source-bound review and never replays the call (4049.2916ms) ✔ changed enabled MCP programs can be explicitly disabled without launching or trusting their changed source (1530.0726ms) ✔ stdio preserves split UTF-8 frames and final replies drained after process exit (2010.9885ms) ✔ server-initiated model requests are refused instead of inheriting provider privileges (756.1336ms) ✔ invalid or over-budget protocol invalid closes only its owned process and is not retried (907.9047ms) ✔ invalid or over-budget protocol oversized closes only its owned process and is not retried (899.5396ms) ✔ cancel and timeout do not resend a dispatched request and pre-cancel has no request trace (1066.1947ms) ✔ startup failure remains undispatched and has no process PID to terminate (508.7942ms) ✔ REA launch filters a fixture credential canary and rejects changed entry identity before a new connection (932.4176ms) ✔ schema1 connections remain readable without inventing observed capabilities or migrating original state (654.5691ms) ✔ protocol, unsupported capabilities and stream choice persist; client claims cannot forge verified evidence (617.3365ms) ✔ ability evidence belongs to exact connection and credential revision; unknown and unsupported reject required routing (0.2858ms) ✔ invalid declarations and inconsistent stream settings reject before registry, revision and credentials mutate (613.0588ms) ✔ model registry keeps original credentials/endpoint intact, rejects stale edits, and forgets only volatile keys on restart (828.6643ms) ✔ config import previews exact revisions, keeps secrets volatile, syncs only when enabled and preserves manual edits on conflicts (808.8426ms) ✔ real HTTP pool fails over 429 to another model and never repeats completed tools; attempts seal actual request bytes and Run identity (3936.6725ms) ✔ pool rotation and weights select members on new Runs; disabled and missing credentials are excluded without requests (5926.737ms) ✔ 401 blocks the credential revision; updating the key restores availability without leaking error echoes (1583.2206ms) ✔ queued request cancellation and in-flight disable release only owned leases and do not dispatch another HTTP request (766.6438ms) ✔ compiled model-center HTTP API persists public profiles, blocks forged origins, and local operations never invoke a model (1963.8624ms) ✔ model registry disk failures restore routing, pool profiles and credential references without partial updates (651.3252ms) ✔ model selection and saved configuration commit together and failures preserve the original mode and endpoint (756.8667ms) ✔ configuration imports roll back the entire batch and retain one usable preview after validation or commit failure (611.7387ms) ✔ HTTP quota retains native code/type/param/request id, bounded original and allowlisted headers, never repeats that connection (1708.6833ms) ✔ Retry-After above 90 seconds remains exact and a short local budget cannot trigger an early retry (1.9102ms) ✔ Retry-After date, expired date, huge value and invalid header retain explicit semantics (1.1467ms) ✔ unclassified 429 remains unspecified and every bounded actual retry retains its own original error (2208.0629ms) ✔ oversized HTTP error saves exactly the bounded prefix and labels its incomplete range (1984.9474ms) ✔ Chat SSE error preserves upstream fields, original failed event, partial tool and reported usage without executing or retrying (2213.5901ms) ✔ Responses failed SSE retains original event/response identity and usage (1966.9414ms) ✔ Responses incomplete SSE retains original event/response identity and usage (2296.0916ms) ✔ unsupported successful Responses output retains original JSON and usage with a local protocol origin (2091.8233ms) ✔ HTTP 200 native error is still an upstream failure rather than a fabricated successful answer (0.9877ms) ✔ JSON-escaped credential echoes are redacted in the original preview and capture as well as parsed fields (0.9322ms) ✔ quota failover uses another independent pool member and never replays a failed stream (2163.7126ms) ✔ native parameters are isolated by pool member, preserve Responses state and bind the original output reservation (0.3202ms) ✔ unsupported native fields and invalid values reject before dispatch instead of silently dropping or overriding constraints (0.3089ms) ✔ native settings for legacy and registered profiles persist without changing credentials or old records and reject stale edits (1362.7298ms) ✔ compiled native settings API saves/restores exact parameters and rejects reserved fields without mutating the profile (2647.9021ms) ✔ legacy connection test sends its saved native fields and selected output limit field without changing the configuration (3086.6562ms) ✔ identity distinguishes missing, invalid, mixed, exact and explicitly declared aliases without guessing (1.1688ms) ✔ chat-completions JSON rejects a different model before tool dispatch, retaining source and reported usage (2716.1262ms) ✔ chat-completions SSE rejects a different model before tool dispatch, retaining source and reported usage (2976.2807ms) ✔ responses JSON rejects a different model before tool dispatch, retaining source and reported usage (1594.6615ms) ✔ responses SSE rejects a different model before tool dispatch, retaining source and reported usage (1505.5379ms) ✔ record policy accepts missing identity but keeps it explicit, and strict declared mapping is frozen in the consumption (1.9178ms) ✔ mixed streamed identities cannot pass strict mode, and interrupted streams retain observed identifiers without authenticating them (1.5938ms) ✔ identity refusal never fails over to another pool member and raw response model cannot leak known credentials (1821.847ms) ✔ strict refusal remains visible even when the same original response exhausts the budget (1407.2267ms) ✔ strict identity crosses the controlled engine bridge with its original source and zero tool dispatch (12600.6674ms) ✔ frozen direct connection policy and delivery reports preserve response identity without filling historical gaps (2710.7742ms) ✔ saved engine budgets verify all Chat/Responses route variants and reject a mismatched output allowance (700.7559ms) ✔ owned CC Switch SQLite imports the selected TOML Responses wire_api and leaves unknown clients untouched (970.372ms) ✔ Responses executes one original tool through Runtime, carries encrypted Items, reports raw usage and seals exact wire bytes (2442.6344ms) ✔ mixed-protocol pool keeps both actual variants and changes route on 429 without re-executing completed tools (3902.3017ms) ▶ Responses incomplete, failed, missing call identity and unsupported hosted outputs never dispatch a tool ✔ Reject {"body":{"id":"resp_owned","status":"incomplete","output":[{"type":"function_call","id":"fc_partial" (1721.5978ms) ✔ Reject {"body":{"id":"resp_owned","status":"failed","output":[{"type":"function_call","id":"fc_failed","cal (1775.2078ms) ✔ Reject {"body":{"id":"resp_owned","status":"completed","output":[{"type":"function_call","id":"fc_bad","cal (1595.9199ms) ✔ Reject {"body":{"id":"resp_owned","status":"completed","output":[{"type":"computer_call","id":"not-supporte (1467.6809ms) ✔ Responses incomplete, failed, missing call identity and unsupported hosted outputs never dispatch a tool (6561.4487ms) ▶ invalid JSON in a nonstream Chat or Responses tool group prevents every tool and pool replay ✔ chat-completions (1702.7535ms) ✔ responses (1651.9808ms) ✔ invalid JSON in a nonstream Chat or Responses tool group prevents every tool and pool replay (3355.2301ms) ✔ wire adapter preserves tool IDs but never passes Responses-private Items to Chat or another endpoint (0.2027ms) ✔ connection JSON import accepts explicit Responses protocol and continues to reject unknown protocols without reading client sessions (553.7614ms) ✔ reservation uses actual known usage from the same route, never lowers estimates or rewrites original receipts (2.6405ms) ✔ reservation verification binds original Task/Run consumption and rejects changed usage or foreign provenance (0.7746ms) ✔ known high input calibrates the next request and stops before dispatch while preserving the completed read (1772.3878ms) ✔ one reported overrun stops legacy requests and keeps actual use and excess (1483.1177ms) ✔ one reported overrun stops pool requests and keeps actual use and excess (1785.1105ms) ✔ one reported overrun stops connection requests and keeps actual use and excess (2093.9882ms) ✔ reported usage below the limit prevents an unaffordable legacy retry before another dispatch (1534.2998ms) ✔ reported usage below the limit prevents an unaffordable pool retry before another dispatch (2209.571ms) ✔ native bridge seals calibrated refusal with original consumption and preserves versioned request verification (13619.2595ms) ✔ original R v1/v2 stay verifiable while v3 excludes the isolated input outlier and keeps the ledger intact (5.1165ms) ✔ v3 never excludes a repeated high ratio, the latest receipt, or a short history (3.0313ms) ✔ reservation proofs reject reordered latest high receipts, altered history and duplicates while old snapshots survive later consumption (3.6671ms) ✔ input-ratio exclusion preserves independently reported output maxima without rewriting historical v2 (1.9737ms) ✔ capability routing skips text-only members before acquiring a lease and records exact refusal in the original Run (2317.1756ms) ✔ all incompatible members reject with no network request, no lease, preserved Task and explicit non-dispatched ledger (1482.2797ms) ✔ explicit synthetic tool probe verifies only actual matching call, never executes it and loses verification on revision changes (1140.4986ms) ✔ plain text probe cannot verify tools; wrong tool nonce and client supplied evidence cannot claim verified support (1384.6856ms) ✔ a disabled connection invalidates an in-flight probe even when toggle mutates its original registry object (899.6164ms) ✔ selected connection owns the actual endpoint, credential, wire model and mismatch even when legacy is configured for GPT (3925.4528ms) ✔ budget-only API and demo/real selection preserve routes, reject injected endpoint fields and publish authoritative selection without keys (3499.9279ms) ✔ another session cannot change default model, mode or budgets while a request is active (5578.9415ms) ✔ HTTP download preserves original malformed SSE bytes and artifact SHA instead of decoding invalid UTF8 (2745.0242ms) ✔ terminated SSE adapts empty optional tool arguments, retains the exact wire bytes and executes the validated batch once (5524.2544ms) ✔ reported SSE usage survives rejected required arguments "" without dispatching any tool (2492.2931ms) ✔ reported SSE usage survives rejected required arguments "{\"path\":" without dispatching any tool (2140.0483ms) ✔ reported SSE usage survives rejected required arguments "null" without dispatching any tool (2418.7367ms) ✔ a stream without DONE retains reported usage but cannot adapt or execute even an optional tool (2404.9917ms) ✔ a timeout after reasoning-only fragments creates one unknown receipt and never invents reported tokens (6028.1478ms) ✔ real Chat SSE carries UTF8 and split tool JSON; Runtime executes only after DONE, stores partial progress and original usage (6449.3747ms) ✔ Responses typed SSE validates original item identity, terminal data, raw usage and tool result pair (3313.3115ms) ✔ incomplete network stream retains unknown attempt and never automatically switches pool or dispatches fragment tools (2465.4224ms) ✔ stopping an active HTTP stream closes only original request and keeps partial text without executing tools (2305.8716ms) ✔ explicit streaming probe uses actual SSE despite unknown declaration and binds its own original attempt (961.1386ms) ✔ coalesced initial SSE events publish the latest partial state even when the upstream stays idle (2598.6439ms) ✔ stream timeout closes the original HTTP response, preserves redacted capture and never retries (1534.6782ms) ✔ stream idle timeout preserves incomplete tool arguments, records unknown use and never fails over (1432.8411ms) ✔ SSE accepts CR-only final event boundaries and retains exact invalid UTF8 bytes without dispatch (0.7718ms) ▶ stream parser rejects truncation, finish length, malformed JSON, duplicate sequence, changed final item and oversized data ✔ reject 0 (0.2399ms) ✔ reject 1 (0.1678ms) ✔ reject 2 (0.1327ms) ✔ reject 3 (5.138ms) ✔ reject 4 (0.2029ms) ✔ reject 5 (0.4543ms) ✔ stream parser rejects truncation, finish length, malformed JSON, duplicate sequence, changed final item and oversized data (6.814ms) ✔ compiled host retains interrupted HTTP 200 and explicitly resumes after the recorded cooldown (15257.8656ms) ✔ compiled host retains interrupted HTTP 503 and explicitly resumes after the recorded cooldown (11073.5741ms) ✔ owned IPv4 service works after a failed IPv6 candidate without a second model POST or global DNS change (25.8611ms) ✔ environment HTTP proxy tunnels the owned request and NO_PROXY sends a separate direct request (10.6306ms) ✔ original connect timeout has a bounded nested cause and does not masquerade as response generation timeout (0.737ms) ✔ POST accepted then reset is unknown, is charged once and never automatically retried (3.5143ms) ✔ partial HTTP bodies retain original request id and bytes, and an unknown pool attempt never switches model (2704.8776ms) ✔ response wait timeout after an accepted POST is unknown; cancellation closes the owned socket and permits explicit reuse (150.674ms) ✔ older configurations receive bounded wait defaults without changing the original object (0.9997ms) ✔ active stream survives sixty seconds; byte activity resets idle but cannot extend total deadline (0.8526ms) ✔ silent stream has its own accurate deadline; caller cancellation and cleanup preserve their cause (1.3523ms) ✔ waiting for headers is bounded independently of stream-idle time (0.2239ms) ✔ wait settings persist, pin new runs and leave model pool, credentials and old run snapshots intact (1750.7594ms) ✔ plain endpoint checks only TCP, records family coverage and never reads model credentials or writes routing/health/attempts (1682.1102ms) ✔ local DNS failure is separate from TCP/TLS and arbitrary input, stale revision and TLS overrides are rejected (632.0645ms) ✔ refused TCP socket preserves its stage and cannot masquerade as a TLS or API response (1420.425ms) ✔ TLS validation succeeds for a trusted owned certificate and rejects the same untrusted certificate (3583.0544ms) ✔ a TCP peer that never answers TLS is bounded, and cancellation waits for owned sockets to close (2009.1307ms) ✔ a stalled lookup is cancellable and the overall deadline bounds DNS without late socket work (1135.1931ms) ✔ dual-stack fallback and explicit family results remain separate and never change global DNS defaults (1452.3624ms) ✔ proxy settings are reported as scope only; diagnostics do not contact the proxy or leak its credentials (1085.3808ms) ✔ history is bounded, survives reopening, marks interrupted work and reports changed revisions without rewriting frozen targets (2302.0013ms) ✔ HTTP API, host activity, configuration edits, model start and upgrade honor the same active diagnostic and stop operation (1773.908ms) ✔ provider shutdown cancels diagnostics and permits explicit reuse after a failed host stop (1433.3446ms) ✔ unchanged proposals cannot hide repeated original-log reads or identical session inspections (11477.9036ms) ✔ reusing a verified proof never resets identical-state detection or creates a new proof (8637.093ms) ✔ coverage merges adjacent pages, separates changed sources and never claims old Run coverage in a continuation (0.7498ms) ✔ paged inspection finds a corrupted old artifact outside the recent fifty and retains original evidence without model/tool execution (5756.4132ms) ✔ inspection rejects ambiguous/out-of-range pagination and reports unknown tool/task identity without resolving it (1531.2294ms) ✔ original source locks all 60 byte-identical files including modules outside the nine-file prompt experiment (1511.4741ms) ✔ a new original executable or missing license cannot silently join the fixed trusted-source profile (1641.6959ms) ✔ original mode has a separate source identity and does not accept retry or privilege fields (1334.6641ms) ✔ actual original model input requires the complete postPrompt once in system context and cannot borrow a user or tool copy (0.3189ms) ✔ Windows directory permissions normalize overlapping grants without allowing an unrelated sibling (1142.3668ms) ✔ durability broker rejects foreign authentication, outside paths and wrong file identity before real fsync (1023.0978ms) ✔ exact patches preserve BOM/CRLF, unicode and all unselected bytes and back up the original on apply (1515.6829ms) ✔ several patches are matched against one original snapshot in any input order and keep LF (1345.045ms) ✔ mixed line endings require exact anchors and preserve untouched formatting (848.1294ms) ✔ ambiguous, overlapping, missing and malformed patch anchors never create a proposal or write (780.8941ms) ✔ stale original hashes and changes before apply preserve owner edits and mark conflicts (705.7007ms) ✔ no-op BOM patches/full content produce no change and replacements cannot split Unicode bytes (574.5645ms) ✔ real owned Vite starts only after approval, records OS ownership and live changes, and stops its tree (10589.553ms) ✔ occupied ports and approval-time conflicts never take over an unrelated listener (6603.5042ms) ✔ background preview final-log failure retains unknown evidence without an unhandled startup promise (10590.2624ms) ✔ stale source, tampered plans, wrong ownership and arbitrary scripts cannot dispatch a preview (12711.4712ms) ✔ restart retains original unknown preview, blocks mutation, reconciles absent identity without replay (6944.5828ms) ✔ an actual shell command keeps Node and UTF-8 output while omitting host-only credentials (844.1261ms) ✔ real npm tests and build do not inherit arbitrary host credentials or Node injection settings (2548.2932ms) ✔ real Python tests and compilation keep their sealed inputs without host credentials (2801.3913ms) ✔ a process exiting between the Windows snapshot and owner query is absent only after a fresh query (1802.3812ms) ✔ owner observation errors remain errors while a process is present, replaced, or cannot be rechecked (1611.1804ms) ✔ cache controls require desktop authority, exact inputs and owned project identity while preserving task evidence (2624.126ms) ✔ cache enablement and derived reuse survive a real host restart without becoming task recovery state (4694.2636ms) ✔ persistent parsing is project/SHA/config bound and produces identical semantic indexes after restart (1534.1064ms) warning: in the working copy of 'README.md', LF will be replaced by CRLF the next time Git touches it warning: in the working copy of 'main.ts', LF will be replaced by CRLF the next time Git touches it warning: in the working copy of 'value.ts', LF will be replaced by CRLF the next time Git touches it ✔ same-size restored-mtime changes, additions, renames and deletions cannot reuse stale entries or import targets (1245.1488ms) Switched to a new branch 'other' ✔ branch and commit changes invalidate even byte-identical sources (3922.0586ms) ✔ corrupt, malformed and schema-invalid records fall back to fresh parsing (944.6124ms) ✔ disable and clear persist without touching sources, sessions, configuration or original artifacts (1862.091ms) ✔ cache capacity bounds disk and memory by bytes and LRU project count (4371.4178ms) ✔ incomplete scans do not load or save cached entries; unavailable files leave the current result scope (1595.2261ms) ✔ project and cache junctions cannot expose or delete an outside directory (710.5191ms) ✔ concurrent file changes refuse a snapshot and clear during pending work cannot repopulate it (881.3984ms) ✔ file access mode changes invalidate parsed material even when bytes are unchanged (673.0932ms) ✔ tracked directory replaced by a junction invalidates access scope and never uses its old cache (1965.5632ms) ✔ JS/TS/React AST indexes actual declarations, exports, imports and verified source lines (897.2442ms) ✔ default expressions, named default exports and declared functions retain their AST evidence (729.1591ms) ✔ incremental refresh invalidates changed/deleted symbols and recalculates reused import targets (1202.1884ms) ✔ Python candidates exclude comments and multiline strings and declare limited evidence (582.8363ms) ✔ indexes isolate project identity and root and evict bounded caches without sharing results (800.8855ms) ✔ Git ignore, credential paths and links stay outside the index and no project files are executed (1385.8451ms) ✔ malformed, binary, non-UTF8 and bounded files expose partial coverage while good files remain searchable (715.5367ms) ✔ index query rejects invalid limits/types and does not imply reference or alias resolution (535.7892ms) ✔ read-only engineering inspection disables configured Git clean filters and preserves index bytes (1764.5701ms) ✔ Python multiline declarations retain candidate scope without interpreting expressions as declarations (922.2956ms) ✔ TypeScript module candidates support import-equals and MTS/CTS while opaque imports stay explicit (697.3401ms) ✔ actual HTTP requests reuse only this project source-checked memory as redacted data and preserve sent evidence after archive/restart (2367.2188ms) ✔ actual HTTP request excludes archived, timed-out and foreign-project memories (4300.2974ms) ✔ actual next HTTP request removes expired memory after a tool observes a source change without replaying that tool (3702.0835ms) ✔ incomplete inventory and excluded-file SHA changes cannot enter actual HTTP memory context or inspect_session snapshots (3082.5517ms) ✔ corrupt original artifact expires its memory and cannot reach an HTTP model request (2623.1578ms) ✔ memory JSON is bounded in escaped UTF-8 bytes and entry count, skips oversized entries and retains complete multilingual values (3921.1929ms) ✔ optional memory yields to the complete goal and original code evidence under the whole-request budget (1272.239ms) ✔ changed sealed memory artifacts or binding fields fail verification while the original request remains immutable (1872.7429ms) ✔ a malformed project declaration omits optional memory while the repair Task still reaches the owned HTTP service (1619.8666ms) ✔ an expiration persistence failure prevents dispatch and is not hidden as an optional scan failure (1345.3277ms) ✔ project memories persist across restart, keep source provenance, and are visible across tasks (1417.1766ms) ✔ source changes expire current project memories and keep the original evidence (1098.6356ms) ✔ incomplete source scans mark current memories unverified without expiring or reusing them (1148.4604ms) ✔ explicit file provenance is rechecked even when the inventory excludes that path (1241.5371ms) ✔ execution snapshots do not expose project memories before their source scope is validated (828.9048ms) ✔ project memory records stay isolated and archive requires the owning session (1171.073ms) ✔ project memory HTTP API exposes read-only source status and explicit writes (2753.2989ms) ✔ two Purge source versions lock nine original inputs without executing JavaScript or postinstall and remain disabled by default (812.3206ms) ✔ incomplete, unsupported, relative and oversized sources cannot enable the experiment (960.2978ms) ✔ invalid source fingerprints and invented privilege or retry fields fail before any parent Task is created (2134.7165ms) ✔ source edits between detection and plan are rejected without creating a parent Task (1697.8019ms) ✔ source edits after sealing cannot launch a child and do not reset its cumulative budget (9388.1698ms) ✔ literal experiment text is sealed into the parent, worker and each new Run while actual host approvals and saved results remain required (31720.7797ms) ✔ source changes after a model response stop before any native tool is accepted, retaining the actual usage once (13319.8519ms) ✔ source changes while approval is pending invalidate the old permission before an engineering check starts (19397.2711ms) ✔ altered text, interpolation, section order and policy flags cannot reuse a sealed experiment (2096.9437ms) ✔ the observed SDK system-message layout and the explicit system field both require exactly one effective section, never a user or tool copy (2269.1082ms) ✔ a missing or duplicated experimental section in the actual input is rejected before any model dispatch (28263.5872ms) ✔ experimental prompt overhead still participates in the original complete-input budget guard (12386.8205ms) ✔ source detection HTTP remains local-authenticated and read-only with no configured model requests (3872.4915ms) ✔ Python HTTP plans keep npm defaults, require owned approval and preserve actual results across restart (9786.7883ms) ✔ an isolated three-module Python Task repairs, checks and delivers a source-bound Git report without extra model calls (44783.7475ms) ✔ Python static inspection does not import project code or probe a project interpreter before approval (1113.3914ms) ✔ a real three-module Python failure, reviewed repairs, tests and compilation retain exact structured evidence (5139.4946ms) ✔ unittest counts repeated execution ordinals and subtest skips rather than deduplicating display names (4289.4564ms) ✔ unittest empty, skipped, expected failure and class initialization errors never verify a task (9270.8109ms) ✔ stdout claims cannot replace actual unittest results and an early exit cannot forge a successful collector (3296.1567ms) ✔ compilation reads actual Python source bytes without executing top-level project code (3753.6419ms) ✔ Node and Python mixed projects retain the existing npm default and require an explicit Python selection (3212.5412ms) ✔ changed project interpreters and forged Python plan bindings execute no check (2311.5264ms) ✔ malformed TOML and invalid Python test paths are reported without expanding npm check scope (1855.2826ms) ✔ stopping a real Python check terminates its child and keeps the completed result as cancelled (2549.5296ms) ✔ real pytest pass, failure, skip, collection errors and teardown errors use collected case outcomes (17376.3004ms) ✔ a real project virtual environment is detected statically and its actual version is recorded only during the approved check (11413.8353ms) ✔ a missing interpreter and repeated pytest item objects cannot create a false verification pass (3419.248ms) ✔ a mixed turn with new bytes and three repeated modules advances, while repeat-only and independent no-op guards still pause (0.9781ms) ✔ runtime sends the next real HTTP request after a W-shaped mixed-progress turn instead of stopping on its trailing repeats (7248.6932ms) ✔ continuation reads advance through historical pages without a false pause while repeated current-Run pages still pause (1.3268ms) ✔ a new tool page replaces old latest-feedback space and still obeys its escaped protocol budget (0.4343ms) ✔ budgeted artifact pages count escaped protocol bytes and reconstruct original multilingual JSON without losing characters (8.6455ms) ✔ known artifact coverage detects semantic repeats despite changing limits, and advancing pages or direct evidence restore progress (0.3673ms) ✔ actual large artifact reads are present in the next bounded HTTP request and next_offset progresses without redoing tools (8588.7803ms) ✔ a real repeated artifact loop pauses the same Task after recorded results, survives reload and only resumes explicitly (10430.8941ms) ✔ actual HTTP continuation after reload sends distinct historically covered pages without false no_progress or replay (9993.334ms) ✔ continuations distinguish unread source, later original line ranges and already returned result wrappers (1228.3474ms) ✔ pure-file batch returns unused small-page space while mixed and historical allocation stays conservative (0.4843ms) ✔ v3 pools mixed file/log reads, serves small sources first and leaves historical/mutating batches unchanged (0.3184ms) ✔ v3 W-shaped mixed reading returns complete small modules and substantially larger first test pages in the next bounded request (6679.1617ms) ✔ observed upstream high input pauses before a new request, preserves reported usage and distinguishes the limiting ledger (977.1898ms) ✔ chat-completions: complete actual Unicode/CRLF pages and small source survive the next bounded wire request (5544.5779ms) ✔ responses: complete actual Unicode/CRLF pages and small source survive the next bounded wire request (5499.6268ms) ✔ anthropic-messages: complete actual Unicode/CRLF pages and small source survive the next bounded wire request (5011.7759ms) ✔ an installed app outside the source drive resolves registered SDK and Purge through actual input verification (1002.7453ms) ✔ standalone registration resolves the upstream metadata directory and missing upstream falls back to the other registered source (918.9392ms) ✔ registered hints keep explicit roots, environment and existing app-relative candidates in their original priority (1565.5776ms) ✔ unknown IDs, duplicate registered IDs, wrong versions and relative or inconsistent metadata are not lookup candidates (1236.0622ms) ✔ oversized, malformed, invalid UTF-8 and non-file registry inputs cannot silently enable a source (1166.8384ms) ✔ a linked registry directory is rejected without reading or deleting its external source (1287.8578ms) ✔ matching research declarations do not replace actual installed version and critical-source checks (1217.4237ms) ✔ runtime guides map archived evidence while all original reports and successful/failed source bytes remain independently verifiable (1531.0933ms) ✔ missing required research reports fail packaging rather than creating a successful placeholder (1325.7602ms) ✔ full verification reports only verified bytes, complete declared scope and actual directory inventory; changed/deleted files still fail (1446.3681ms) ✔ preflight and activation continue one monotonic operation bound to installer hash; terminal replay and changed origin are rejected (9423.147ms) ✔ preflight failures and fixed-candidate mismatches retain the actual reason without publishing a pointer (5310.3884ms) ✔ cancel finishes only the original ready preparation, preserves data and pointer and refuses replay (7149.2255ms) ✔ the installation lock rejects concurrent same-id operations before they can overwrite original progress (4616.1986ms) ✔ original post-publication failure is linked to its rollback receipt and every original byte survives (5264.9695ms) ✔ a candidate mutation at the publication boundary still fails even after an earlier successful full validation (10217.7571ms) ✔ public progress marks a lost process unknown without editing its original record or claiming completion (6349.0196ms) ✔ throwing console/UI progress observers cannot decide the installation result (4762.3876ms) ✔ a progress-only write failure after committed publication returns the authoritative receipt and a warning (4122.9355ms) ✔ browser and MCP approvals use actual tool categories; mobile exposes only category while keeping original fingerprint (553.924ms) ✔ the compiled runtime version constant tracks package.json so a release bump cannot ship a stale served version (8.6483ms) ✔ context APIs retain exact request identity, paginate archived groups and reject foreign sessions/runs (3361.989ms) ✔ delivery reports verify request body, sealed plan and original archived bytes, and show corruption as needing verification (3968.3014ms) ✔ native seed framing reserves capacity before optional history without changing the selected window (1541.3718ms) ✔ direct code evidence remains selected while recent historical-index browsing is archived within the complete request budget (1731.8366ms) ✔ one long user turn becomes bounded complete groups with exact UTF-8 history and recent JSON arguments (3971.4204ms) ✔ tool definitions, exact system bytes, output reserve and remaining Run/Task budget are accounted before requests (881.7259ms) ✔ oversized required user instructions pause intact and can be explicitly resumed after increasing the window (1692.4136ms) ✔ all task instructions, constraints and acceptance remain in the runtime system instead of an eight-item or byte prefix (1116.9479ms) ✔ a newest manageable multi-tool group stays exact even beyond the working target; oversized earlier groups are archived (859.1809ms) ✔ invalid orphan result is rejected before archive or network preparation (719.9476ms) ✔ invalid missing result is rejected before archive or network preparation (967.8696ms) ✔ invalid wrong result is rejected before archive or network preparation (752.4278ms) ✔ invalid duplicate call is rejected before archive or network preparation (736.5484ms) ✔ a group larger than the artifact limit uses verified Unicode-safe chunks without replacing tool JSON (2250.1285ms) ✔ changed or missing archived evidence stops reuse and is never silently recreated (1317.9859ms) ✔ transaction failure prevents publishing a plan and any corresponding model request (1072.4206ms) ✔ the Provider rejects body mutation after a sealed plan before calling HTTP (883.0783ms) ✔ HTTP failure and explicit same-Task resume reuse exact completed results and request hashes without repeating a command (12601.4194ms) ✔ a real process exit after plan persistence marks interruption and never issues an automatic model request (1301.8188ms) ✔ local draft preserves exact original ranges and duplicate source instances without inferring outputs or accepting facts (768.0836ms) ✔ review keeps unknown blocking, references distinct from output and boundaries distinct from compliance evidence (1393.4481ms) ✔ draft adoption must account for every original range and does not fill paths from a reference (1094.9296ms) ✔ source and requirement changes, other Tasks and tampered draft bytes are rejected without losing originals (1990.9217ms) ✔ partial explicit source review leaves the other source unreviewed while old file evidence remains valid (1301.526ms) ✔ bounded draft does not claim full source coverage and cannot acknowledge an omitted remainder (692.5454ms) ✔ draft HTTP preserves its original artifact across restart and rejects stale scope and source versions (4723.2002ms) ✔ requirements HTTP binds selection, enforces optimistic revisions, restores original proofs and leaves provider history untouched (5072.0335ms) ✔ research evidence binds actual original bytes; model completion is not user acceptance (1676.8896ms) ✔ same text with distinct user source instances stays distinct; an appended requirement only invalidates its own revision (1454.7623ms) ✔ foreign project, old Task and fabricated evidence strings are rejected with preserved failed proofs (3253.9241ms) ✔ source changes invalidate prior proofs and cannot certify a read that was already stale at binding (2495.4273ms) ✔ missing or tampered artifacts never become evidence and original failures remain after recovery (1900.5136ms) ✔ checks omitted by explicit agreement are reported as not executed and never passed (1128.684ms) ✔ user confirmation stays separate from engineering evidence, is bounded to current inputs and cannot be silently converted (1349.2194ms) ✔ existing multi-file results keep provenance and invalidate only evidence whose files changed (1767.9853ms) ✔ unrelated additions preserve the checked file facts but remain visible as unreviewed user sources (1708.5016ms) ✔ actual Git preview and publication recheck changed requirement bindings without touching HEAD or the owner index (8749.0856ms) ✔ no-test file application and manual completion keep pending user requirements for explicit review (2241.2265ms) ✔ revision races, forged source ranges, evidence fields and deletion cannot overwrite sealed requirements (1044.033ms) ✔ requirement binding prevents completion and Git acceptance after a review changes; legacy tasks stay explicit (3408.8583ms) ✔ deep Windows data directories keep REA output within the external path allowance and retain owned evidence (3060.2762ms) ✔ user research registration creates an owned Task without model requests and seals the exact artifact identity (1214.9973ms) ✔ complete research binds original Evidence and Bundle, pages exact fields and reuses verified results after reload (2186.366ms) ✔ artifact changes before analysis refuse dispatch and changes during analysis preserve unverified original responses (3049.8943ms) ✔ owned cancellation closes the connection, concurrent and automatic retries are refused, explicit new analysis preserves history (1939.5031ms) ✔ a dispatched transport timeout stays unknown and reload never launches or kills a saved PID (2045.6159ms) ✔ corrupt Evidence parts and changed binding records are refused instead of silently re-analysed (2168.0144ms) ✔ research documents over the single-artifact limit preserve exact Unicode bytes with owned verifiable parts (1129.2927ms) ✔ research tools are only sent for registered Task artifacts and actual HTTP uses the sealed tool contract without replay (5968.592ms) ✔ file identity rejects directory and invalid formats without running a target (858.6699ms) ✔ a stop from the event loop interrupts immediate model replies before later tools (2337.3818ms) ✔ a stop after the first tool preserves a complete interrupted multi-call protocol group (2671.1273ms) ✔ the compatible HTTP loop preserves call IDs, validates errors and waits before executing commands (8041.3713ms) ✔ denial prevents command execution and produces a paired denied result (2508.5775ms) ✔ stopping during approval interrupts pending calls without executing or replaying them (2636.1556ms) ✔ stopping a running fixture command terminates its own process tree and retains uncertain state (3150.6544ms) ✔ command output decodes split UTF-8 and excludes the workbench key from child environment (1116.4883ms) ✔ model-list testing uses a tentative key without changing the saved connection and redacts service errors (6.2523ms) ✔ truncated or repeated model calls fail before execution; tool input validation rejects unknown capabilities (4304.8398ms) ✔ request projection keeps exact tool pairs and complete recent user instructions while preserving disk history (1687.3098ms) ✔ local skill revisions require explicit enable, retain original text, survive restart and stay within the selected project (1338.3291ms) ✔ registration rejects non-SKILL paths, oversized/non-UTF8 files, duplicate sources and wrong field shapes without running project code (879.7803ms) ✔ actual HTTP models receive only enabled current-project metadata, then a redacted owned skill page with complete tool pairing (3780.0937ms) ✔ source edits after a catalog request prevent the read and appear as changed on the next request without replay (3878.4875ms) ✔ disabled and foreign skills cannot be read by guessing their IDs from another project (4259.5565ms) ✔ skill catalog takes remaining escaped JSON budget and cannot evict the user goal or current code (1177.4703ms) ✔ multilingual skill body pages fit actual tool JSON bytes and every page uses the same immutable artifact (5581.2818ms) ✔ registry failures roll back enable and refresh state; corrupted snapshots block requests before network (1887.4324ms) ✔ saved skill catalogs retain SHA-bound historical evidence and reject corrupt catalog blobs (3167.8788ms) ✔ HTTP skill administration enforces project scope, revision checks and idle mutation without model calls (3364.5769ms) ✔ JSON state keeps an atomic backup and preserves corrupt primary files (1179.1444ms) ✔ tool-pair recovery fills missing results and drops orphan or duplicate results (0.332ms) ✔ restart pauses unfinished work, retains proposals and never replays a command (1335.3427ms) ✔ public snapshots carry increasing revisions without the private provider transcript or config key (1058.5131ms) ✔ a checkpoint seals all task requirements and original results without a model request and explicit continuation keeps the same Task without replay (5065.0913ms) ✔ checkpoint sources, requirements, plan and original tool changes block continuation and leave original checkpoint bytes intact (3543.7001ms) ✔ unknown tool effects and pending proposals stay blocked in recovery packages rather than being changed into completed work (4298.1207ms) ✔ checkpoint and referenced result corruption are rejected before dispatch and historical blobs are never rebuilt (4025.5459ms) ✔ checkpoint ownership, exact confirmation and concurrent operations reject foreign or stale resumes (4053.0726ms) ✔ Git identity changes and altered plan proofs cannot be accepted as the sealed recovery state (4422.0629ms) ✔ checkpoint records survive SQLite reopen and do not trigger automatic model or tool replay (4311.1783ms) ✔ incomplete scans and in-progress project activity refuse checkpoint creation while private credentials are redacted from saved requirements (4191.6705ms) ✔ oversized recovery evidence is refused before reading or publishing the checkpoint (3488.8853ms) ✔ compiled checkpoint HTTP APIs bind current Task/Run, isolate sessions and only resume on an explicit matching confirmation (9277.4752ms) ✔ actual process exit before checkpoint commit retains only durable records without replay (4002.2412ms) ✔ actual process exit after checkpoint commit retains only durable records without replay (3969.4571ms) ✔ a real two-reply failure replay reaches request three at 16K with complete goal, original sources and sealed tool evidence (9554.2885ms) ✔ request deduplication retains long original intent, extra requirements, constraints and acceptance without changing the stored contract (2843.6723ms) ✔ only the routine context avoids plan echoes; diagnostic API and requests without an included plan keep original observations (960.8493ms) ✔ operational questions are diagnostic while explicit repair requests remain execution instructions (1.0276ms) ✔ an explicit prohibition on running engineering checks is not interpreted as a requirement to run them (1.7016ms) ✔ file delivery requires a positive instruction, not saved evidence or a prohibited write (0.2714ms) ✔ an unsubstantiated text claim cannot complete a requested file delivery (1651.4261ms) ✔ HTML without tests ends at reviewed file delivery and never offers automatic rechecking (3457.491ms) ✔ diagnostic answers retain the unfinished task, limits and check result, and enforce read-only tools (4866.538ms) ✔ review continuation is a system action and adds no testing demand or fake user instruction (3734.1997ms) ✔ a provider cannot execute tests against an explicit no-test task even if it ignores advertised tools (2876.6852ms) ✔ a budget pause becomes visible only after its input snapshot can be sealed with the original Run (2093.2521ms) ✔ alternating duplicate file reads and listing files still pauses without treating new pages as duplicates (19749.852ms) ✔ task-plan HTTP mutations bind current Task/Run and retain original evidence through real host restart (5654.0971ms) ✔ concurrent stale HTTP plan revisions cannot replace conditions or attach proofs to a different plan (3162.0132ms) ✔ a real escaped long-file tool message keeps its proof identity within a tight request budget (1285.6589ms) ✔ task plans reject fabricated completion, invalid scopes and cyclic or missing dependencies (0.5019ms) ✔ sealed plans retain goal, contract hash and original steps through SQLite reopen without executing tools (666.9819ms) ✔ proofs require original owned reads, honor dependencies and reject altered invocation arguments and bytes (1504.7431ms) ✔ new user requirements invalidate current proofs and reseal the same original operations in a new revision (1388.0933ms) ✔ applied change-set evidence binds reviewed current files and invalidates dependent steps after outside edits (2283.166ms) ✔ actual npm checks are verified by original counted logs and current source, with failure history retained (7891.078ms) ✔ real desktop and mobile browser proofs bind required actions and PNGs, rejecting source changes, a later failed plan and changed builds (12562.1533ms) ✔ Git delivery blocks incomplete plans and changed preview bindings, preserving HEAD and owner index before a fresh verified commit (15235.0864ms) ✔ the runtime supplies owned read IDs and completes a plan only after original step evidence verification (4574.8662ms) ✔ premature model completion pauses the original Task and resume preserves the existing plan without replay (4319.5666ms) ✔ plan operations participate in project activity and orderly shutdown (707.7292ms) ✔ actual process exit before plan save restores only committed plan evidence without replay (1331.8007ms) ✔ actual process exit after plan save restores only committed plan evidence without replay (1572.5009ms) ✔ actual process exit before proof save restores only committed plan evidence without replay (1580.3526ms) ✔ actual process exit after proof save restores only committed plan evidence without replay (1532.6661ms) ✔ the latest oversized session result reaches the next request together with already read source modules (2041.3016ms) ✔ every call in a large latest group keeps its actual outcome, including denied and unknown results (1553.5228ms) ✔ a small complete latest tool group is sent once and needs no duplicate feedback view (1138.4076ms) ✔ real HTTP repeated state inspection pauses after three identical engineering outcomes and preserves tool pairing (4110.3781ms) ✔ fixed normal exploration, pagination, SHA changes and repeated ranges are external observations, not a score (1.2998ms) ✔ read-only research is valid work; model summaries and unchanged timestamped listings do not become engineering evidence (0.3764ms) ✔ new permissions, network and broken tools are failure evidence; repeated invalid inputs remain candidates (0.3856ms) ✔ waiting for original long checks, review and approvals stays waiting across repeated observations (0.4815ms) ✔ baseline facts, real repairs, current-source check results and requirement evidence have distinct growth (0.2111ms) ✔ bounded memory keeps unknown instead of claiming an unretained range is repeated; Task/Run instances cannot mix (12.3188ms) ✔ actual runtime records growing pages then a repeated page without changing model rounds, stop rules or original tool pairs (4407.0144ms) ✔ actual invalid tool calls repeat under the original step budget; observation never adds an early stop (3626.0275ms) ✔ actual command approval is observed as waiting and denial stays a permission failure (2719.1013ms) ✔ network errors remain model failures with shadow failure evidence, never a fabricated no-progress cause (4552.7798ms) ✔ observer receipt failure cannot change tool results, model budget or completion (1967.8721ms) ✔ a later receipt recovers after an observation write failure without misaligning retained source receipts (2909.4723ms) ✔ real applied files, counted checks and current requirement proofs grow after an existing-facts baseline (4869.2003ms) ✔ receipt retention stays aligned at the cap and a foreign step cannot enter the Run (3679.0578ms) ✔ an actual approved long check records its owned running job as waiting before the original counted result (7739.275ms) ✔ a live SQLite host with the actual 1.9 artifact object upgrades, keeps its original blobs and verifies real browser PNG bytes (5325.7741ms) ✔ failed upgrade restores the original artifact object, provider and request handler (2206.6242ms) ✔ claiming the same runtime owner twice preserves its original metadata and every database row (1337.6764ms) ✔ sessions without controlled-engine history retain their prior export shape and never expose another session record (1134.6946ms) ✔ live workers, controlled drivers, or MCP operations refuse an upgrade before changing handlers, managers or database rows (606.3543ms) ✔ an idle upgrade loads the newly introduced engine record map, resets obsolete managers and preserves original rows, sessions and provider objects (709.4397ms) ✔ an installation failure restores old managers, original record maps, handlers and all existing database rows (1732.838ms) ✔ the owned controlled driver remains an active session and project even after outer task activity has ended (949.1529ms) ✔ release and upgrade help or invalid arguments have no file or network setup side effects in an independent directory (1116.7019ms) ✔ hot upgrade adopts a retained virtual data root without reopening SQLite or changing rows, Provider or routing (1598.2965ms) ✔ failed adoption restores the original virtual data-root name and original open database (924.5442ms) ✔ hot upgrade refreshes retained model and importer methods while preserving credential and routing objects (915.6222ms) ✔ hot upgrade refuses active probes and restores retained model methods if installation fails (596.5812ms) ✔ the upgrade compatibility gate includes the served current version and the previous formal patch (0.409ms) ✔ a directly started compiled host upgrades twice with the same PID, sessions, applied bytes and volatile provider (4034.1526ms) ✔ the running 0.1.0 process upgrades in place and retains its project, session, files and volatile model connection (7094.8168ms) ✔ HTTP worker identities require a current sealed plan, reject overrides and close unused plans without touching source (19260.4398ms) ✔ HTTP unknown recovery only accepts the original evidence SHA and preserves changed source for explicit retention (24008.6373ms) ✔ verified isolated outputs form a new parent batch, rerun real checks and commit only reviewed source (64192.7882ms) ✔ overlapping worker paths are explicit conflicts even if each independent task checks successfully (36688.5032ms) ✔ source edits after isolated work preserve the owner file and invalidate collection without proposing a merge (33114.8232ms) ✔ changed child bytes after parent preview block the entire batch before any parent write (53297.8407ms) ✔ corrupt original check logs cannot become collected outputs or a parent review (40249.0736ms) ✔ stopping a parent accounts late responses and executes none of their proposed tools (16245.7358ms) ✔ unreported interrupted requests reserve estimated usage and the total deadline cannot reset on resume (16003.274ms) ✔ a deadline before worker request dispatch records no invented usage and cannot reset on resume (15999.9026ms) ✔ a response exceeding the cumulative token allocation stops before registering or executing its tools (15673.8625ms) ✔ cumulative tool allocation rejects the entire returned batch before any tool side effect (14971.1908ms) ✔ single response 16 tool ceiling rejects the entire returned batch before any tool side effect (16029.2994ms) ✔ consumed request steps remain consumed across explicit worker resumes (20191.1898ms) ✔ a saved completed response with missing usage is conservatively accounted exactly once during recovery (16391.464ms) ✔ an unexecuted second workspace plan resumes its original identity without recreating the first checkout (19530.9241ms) ✔ a verified interrupted parent proposal resumes sealing without duplicating or replacing its original change (60487.6054ms) ✔ recovery attaches a sealed parent batch saved before its group reference without creating another batch (55273.5028ms) ✔ explicitly retaining an unexecuted parent merge cancels its batch and frees the next parent goal (51101.3994ms) ✔ unknown collaboration blocks ordinary proposals but permits only original batch recovery and rollback (57745.5032ms) ✔ after a restart the batch application lazily reinstates the worker result guard before any parent write (57832.1433ms) ✔ unknown original tool inputs must match before acknowledgement, and reviewed checkpoint IDs preserve the unknown fact (22285.7028ms) ✔ a parent stop observes and waits for a manually approved child check and its real descendant process (20199.7137ms) ✔ a real host crash preserves unknown checks, refuses a still living owned PID and never replays a model or tool (32916.9922ms) ✔ a persisted child initialization resumes the original control run after a failure before its worker record (15258.8816ms) ✔ failed request intent persistence never dispatches a worker request or changes source (15346.6696ms) ✔ failed usage persistence rolls back the uncommitted usage and conservatively reserves one unknown request without tools (15114.4233ms) ✔ a child file change after observation prevents acknowledgement even when the tool inputs did not change (14761.7655ms) ✔ worker scopes and allocated budgets reject aliases, extra permissions and aggregate overflow (557.2358ms) ✔ two sequential Git workspaces freeze a new parent goal, separate owned tasks and preserve all source bytes (16342.2706ms) ✔ real isolated changes, approvals and npm checks produce original owned outputs, not model completion claims (43640.1946ms) ✔ forged worker tools cannot execute terminal or Git even when absent from the advertised tool list (15980.5768ms) ✔ a completion-only model pauses each worker without fabricated changes or checks (19114.0113ms) ✔ restart retains frozen scope and marks interrupted parent and workers unknown without sending requests (16559.4544ms) ✔ all actually returned same-version fragments reach the bounded request, including overlap and Unicode, with independent original provenance (3974.1422ms) ✔ unread gaps and different range origins remain separate; partial sending names the precise missing already-observed range (936.9716ms) ✔ same-SHA conflicting overlaps and invalid coordinates fail closed, while a new SHA never inherits old fragments (1107.3078ms) ✔ versioned old views keep the original single-page semantics and verification (802.2804ms) ✔ small code and actual failed command stay visible when their original multi-tool group cannot fit (1371.2854ms) ✔ mixed group retains six distinct small file observations with exact multilingual text and bounded request bytes (1697.0296ms) ✔ working observations are bound to the current Task and stop treating a pre-application file revision as current evidence (2234.886ms) ✔ a sealed observation rejects changed or missing original result artifacts instead of recreating evidence (1298.4362ms) ✔ a real passed-check observation is excluded after a source edit invalidates the check, while its historical sent evidence remains valid (5236.9947ms) ✔ budgeted file pages reconstruct UTF-8 text including long single lines and surrogate boundaries (4.6201ms) ✔ actual HTTP multi-tool reads survive an oversized runtime snapshot and the sealed evidence survives reopening (10838.9701ms) ✔ actual HTTP file pagination advances over a long file with each returned page visible in the next request (8784.405ms) ✔ later mixed history pages cannot displace the actual module sources and complete requirements (6422.3906ms) ✔ a later narrow page of the same file revision cannot evict the wider actual observation (3384.9257ms) ✔ six small source files stay complete across metadata turns even when mandatory evidence exceeds the soft target (3718.6371ms) ✔ historical read-artifact navigation points to the original artifact and next page, never its result wrapper (1561.911ms) ✔ a 1.6.1 full-provenance sent view remains verifiable after introducing the compact view (1625.8229ms) ✔ actual HTTP requests keep four observed modules after a later large history read and preserve all source bytes (18343.9793ms) ✔ workspace HTTP ownership, reviewed creation and reopen survive restart without changing source files (13750.4626ms) ✔ isolated model Task applies a batch, executes real checks and commits at the Git review boundary with a source-bound report (42168.8597ms) warning: in the working copy of 'src/空 格.mjs', LF will be replaced by CRLF the next time Git touches it ✔ worktree checkout disables hooks, fsmonitor and all filter command variants while honoring built-in EOL conversion (8803.1152ms) ✔ an ignored external file at checkout is observed and never accepted as pristine creation (7292.8986ms) ✔ outside content hidden by assume-unchanged cannot pass workspace verification and blocks further source execution (5625.9948ms) ✔ workspace preparation refuses known shared-repository activity and preserves archived child sessions (7519.886ms) ✔ compact checkout identities create in deep Unicode source directories while retaining the human branch name (7272.163ms) ✔ Windows Git metadata path limits reject creation before publishing a branch or changing source bytes (2498.3073ms) ✔ a pending source snapshot participates in shutdown and completes before storage can close (3777.2609ms) ✔ source configuration values stay private and a configuration change invalidates a saved review (3186.5027ms) ✔ workspace review rejects excluded dependency before any checkout (3010.724ms) ✔ workspace review rejects large committed blob before any checkout (2832.0531ms) ✔ workspace review rejects case collision before any checkout (3063.1484ms) ✔ workspace review rejects committed symbolic link before any checkout (3375.2629ms) ✔ workspace review rejects external config include before any checkout (2961.7637ms) ✔ actual process exit at workspace intent preserves source and reconciles without creation replay (5218.8592ms) ✔ actual process exit at workspace git_started preserves source and reconciles without creation replay (6497.8491ms) ✔ actual process exit at workspace git_finished preserves source and reconciles without creation replay (7209.5876ms) ✔ actual process exit at workspace verifying preserves source and reconciles without creation replay (6819.4505ms) ✔ actual process exit at workspace registering preserves source and reconciles without creation replay (8611.3051ms) ✔ actual process exit at workspace project_registered preserves source and reconciles without creation replay (8138.8653ms) ✔ actual process exit at workspace session_registered preserves source and reconciles without creation replay (8712.9101ms) ✔ actual process exit at workspace ready preserves source and reconciles without creation replay (6796.094ms) ✔ outside changes at project registration remain unknown and preserve the already registered child identity (7461.6161ms) ✔ managed worktree preserves source index and dirty files, opens a distinct project and binds Run identity (6471.0658ms) ✔ stale review, branch/path collisions and forged ownership do not run worktree creation (3046.0131ms) ✔ workspace project verification is read-only and rejects physical Git identity changes (4913.0754ms) ℹ tests 899 ℹ suites 0 ℹ pass 898 ℹ fail 1 ℹ cancelled 0 ℹ skipped 0 ℹ todo 0 ℹ duration_ms 2093116.1794 ✖ failing tests: test at tests\browser-api.test.mjs:37:1 ✖ self harness browser tools require approval and feed verified actual page fields back to the original Task (16110.9832ms) AssertionError [ERR_ASSERTION]: 模型服务暂时无法处理请求(HTTP 500)。 The expression evaluated to a falsy value: assert.ok(latest.content.includes('已完成')) + actual - expected + 'failed' - 'completed' at TestContext. (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/tests/browser-api.test.mjs:59:10) at process.processTicksAndRejections (node:internal/process/task_queues:104:5) at async Test.run (node:internal/test_runner/test:1409:7) at async Test.processPendingSubtests (node:internal/test_runner/test:974:7) { generatedMessage: false, code: 'ERR_ASSERTION', actual: 'failed', expected: 'completed', operator: 'strictEqual', diff: 'simple' }