✔ Messages builds native system/tool history, bounds images, and seals a distinct mixed-pool variant (8.535ms) ✔ Messages URL and native parameter contracts preserve other protocol paths and hard output budget (0.6341ms) ✔ native usage includes both cache buckets once, retains raw and rejects invalid counters (0.1629ms) ✔ real HTTP Messages JSON executes one tool, retains opaque history, actual URL and cache accounting (3261.1925ms) ✔ real HTTP Messages SSE executes one tool, retains opaque history, actual URL and cache accounting (3075.4012ms) ✔ Messages api-key authentication is explicit; mismatched model remains blocked with original usage (5.2051ms) ✔ Messages bearer authentication is explicit; mismatched model remains blocked with original usage (3.8657ms) ✔ Messages redirect is not followed, either credential stays at the original server (3.5591ms) ✔ strict native SSE reads the nested start model and retains mismatch source without retries (3.0393ms) ✔ a closed native tool block cannot execute before message_stop, then executes exactly once (2820.3363ms) ✔ mixed Chat/Messages pool seals both variants and preserves one completed tool on failover (3778.0575ms) ✔ native JSON rejects incomplete stops, duplicate tools and hosted content (0.3241ms) ▶ strict SSE validates all blocks, terminal, event names, UTF8 and tool JSON before dispatch ✔ reject native stream 0 (1.1216ms) ✔ reject native stream 1 (0.255ms) ✔ reject native stream 2 (0.1729ms) ✔ reject native stream 3 (0.1649ms) ✔ reject native stream 4 (0.1699ms) ✔ reject native stream 5 (0.1694ms) ✔ reject native stream 6 (0.1442ms) ✔ reject native stream 7 (0.1213ms) ✔ reject native stream 8 (0.1797ms) ✔ strict SSE validates all blocks, terminal, event names, UTF8 and tool JSON before dispatch (3.1563ms) ✔ upstream SSE error retains original type/request ID and initial usage, never retries (4.3121ms) ✔ stop cancels the owned stream and retains partial consumption without running tool or replaying (67.0431ms) ✔ saved Messages import preserves auth choice, rejects endpoint duplication, and keeps source file untouched (853.5155ms) ✔ large source receipts use cursor-preserving invalidation without filling the event stream or changing the session (5.1481ms) ✔ small event updates retain the current session and actual cursor (0.1139ms) ✔ the bundled local service enforces host, origin, token and JSON object contracts (2199.8962ms) ✔ an in-flight connection probe prevents configuration changes and another probe until its actual result returns (2358.8747ms) ✔ API review applies verified content, detects conflicts, exports sessions and keeps keys in memory (4963.6565ms) ✔ a real bundle restart preserves proposals, interrupts old approvals and requests fresh consent on resume (10924.2605ms) ✔ runtime diagnostics and model-response validation use actual records and preserve the saved connection (2360.7049ms) ✔ persistent event replay is read-only and keeps run/artifact identity across reconnects (8734.9721ms) ✔ real browser HTTP plans, original field pages and PNG bytes preserve ownership and survive host restart (7781.0963ms) ✔ self harness browser tools require approval and feed verified actual page fields back to the original Task (15343.8579ms) ✔ browser host recovery retains unknown root outcome, blocks replay, and requires an explicit owned acknowledgement (3930.305ms) ✔ late HTTP stop of a finished browser job cannot interrupt a later job in the same Task (7876.5814ms) ✔ actual failed page blocks source-passed delivery; a new passing page permits a bound commit and preserves both original records (19816.1482ms) ✔ a build-only change after Git review invalidates the actual browser evidence and leaves HEAD and the protected index intact (9993.2845ms) ✔ actual DOM controls include validated unique CSS selectors across repeated parent buttons and escaped labels (8363.65ms) ✔ browser field coverage distinguishes sections and Run windows while retaining original source identity (0.4715ms) ✔ repeated original browser pages pause a live Task without re-opening the browser or breaking tool pairing (6778.9127ms) ✔ browser plan rejects outside builds, invalid interaction scope and unbounded steps before execution (0.8915ms) ✔ real owned Edge interactions, assertions, raw PNG and stopped process survive exact result reuse and host reopen (4039.2911ms) ✔ failed real page assertion preserves actual value and forbids silent retry of the original plan (9284.2329ms) ✔ build/source mutation and sealed step tampering reject before browser dispatch; earlier task cannot execute plan (2852.4759ms) ✔ outside network never reaches another local service, and original JavaScript/console failures remain failures (4110.8937ms) ✔ stopping a real in-flight browser preserves finished steps, closes its owned root and never replays (4798.0503ms) ✔ 原PNG损坏使原结果校验失败,不能以已有通过状态复用 (4560.5423ms) ✔ fixed browsing contracts reject local host APIs, credential URLs, protocols, cross-origin and arbitrary execution (1.0016ms) ✔ attested desktop queue preserves same-tab observations, original PNG bytes and approval-bound one-shot actions (1744.0776ms) ✔ page data cannot grant domains, interaction or other Task scope; nonce and ownership remain fixed (1842.991ms) ✔ unknown dispatched submission blocks even after a new observation, rejects late success and retains explicit acknowledgement (1815.996ms) ✔ downloads require exact approved address, remain original binary artifacts, and cannot reuse consumed observation (1480.4399ms) ✔ revoking shared persistent profile closes all its tabs and clears only that profile after original desktop acknowledgement (2494.7678ms) ✔ service recovery never reopens old tabs or replays dispatched actions and keeps unknown original receipts (2575.5445ms) ✔ agent browser flow waits for original Task/Run approval, dispatches once, and returns paired original results (8736.3693ms) ✔ changed approved input and stopped dispatched agent action cannot be accepted or replayed (2786.3553ms) ✔ browser inventory gives bounded metadata and original references instead of repeating full website histories (2272.1653ms) ✔ receipt storage failure settles the original dispatched action as unknown and never hangs or accepts late success (1787.8538ms) ✔ input that fits the engine window but exceeds remaining credit diagnoses the credit limit (0.4943ms) ✔ window overflow and historical pauses retain distinct and honest diagnoses (0.0794ms) ✔ R observed reservation shortage keeps the window-fit diagnosis separate from Run/Task credit (0.0938ms) ✔ a budget pause persists concrete text/tool/history/image breakdown and never dispatches an oversized request (6791.0088ms) ✔ explicit frozen preview resumes the same Task in a new Run, preserves completed tools and does not change global config or selected model (7004.0034ms) ✔ insufficient budgets, silently clamped capacity and arbitrary config fields reject without a physical request or mutating task budget (4108.33ms) ✔ changed session, changed registry, expiry and altered frozen preview all require a new explicit preview (4851.8349ms) ✔ preview preserves unknown invocation status and increasing Task allowance is committed only with confirmed new Run (6280.0314ms) ✔ batch APIs enforce ownership and review fingerprints, apply once, and persist the completed batch across restart (9998.9665ms) ✔ delivery reports bind actual batch identity and reject corrupted snapshot evidence without changing project files (5647.3611ms) ✔ a compatible model can seal proposals through the real HTTP loop while application stays behind user review (10295.6929ms) ✔ a sealed batch at the step limit hands off and continues the same Task into actual checks without replaying saved tools (21758.2827ms) ✔ applied batches cannot turn cancellation, other budgets, restart or unknown commands into a review continuation (3136.494ms) ▶ complete sealed batch hands off after all tools settle with no next model request ✔ complete (6633.4934ms) ✔ later-error (7082.6784ms) ✔ complete sealed batch hands off after all tools settle with no next model request (13716.9538ms) ▶ partial, changed, failed or cancelled batches cannot claim complete review handoff ✔ partial (5393.319ms) ✔ new-proposal (7100.977ms) ✔ failed (6950.9569ms) ✔ conflict (10984.6767ms) ✔ cancelled (9096.4512ms) ✔ unknown (6529.6966ms) ✔ tampered (6149.4238ms) ✔ partial, changed, failed or cancelled batches cannot claim complete review handoff (52206.7696ms) ✔ actual process exit at file-batch prepared preserves durable identity and never replays writes (2584.2786ms) ✔ actual process exit at file-batch application_started preserves durable identity and never replays writes (2766.252ms) ✔ actual process exit at file-batch write_intent preserves durable identity and never replays writes (2811.4537ms) ✔ actual process exit at file-batch staged preserves durable identity and never replays writes (3018.469ms) ✔ actual process exit at file-batch replaced preserves durable identity and never replays writes (4621.456ms) ✔ actual process exit at file-batch file_recorded preserves durable identity and never replays writes (7163.4784ms) ✔ actual process exit at file-batch finished preserves durable identity and never replays writes (5958.1769ms) ✔ actual process exit at file-batch rollback_started preserves durable identity and never replays writes (4357.5397ms) ✔ actual process exit at file-batch restore_intent preserves durable identity and never replays writes (4295.9254ms) ✔ actual process exit at file-batch restore_staged preserves durable identity and never replays writes (4109.4092ms) ✔ actual process exit at file-batch restored preserves durable identity and never replays writes (3800.7983ms) ✔ actual process exit at file-batch restore_recorded preserves durable identity and never replays writes (4054.4312ms) ✔ actual process exit at file-batch rollback_finished preserves durable identity and never replays writes (4225.6225ms) 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 (3237.8421ms) Switched to a new branch 'owner-branch' ✔ a conflict in the last file or Git branch prevents writing every file in the batch (3917.5918ms) ✔ duplicate canonical paths, ancestor collisions, links, dependencies and Windows aliases never form a batch (1656.5664ms) ✔ fingerprints, changed proposals and corrupt snapshot artifacts cannot write or restore any target (2924.8755ms) ✔ a failure after replacement restores only batch bytes and never retries an old application (2278.8217ms) ✔ rollback preserves owner edits, blocks other project writers, and explicit acknowledgement releases recovery (2579.8502ms) ✔ an unattempted owner-created file identical to a suggestion remains external and is never removed (2387.8837ms) ✔ active batch requests are serialized and cancellation leaves a verifiable ended batch (1967.4662ms) ✔ batch application and rollback preserve unrelated Git index and worktree bytes (3525.6481ms) ✔ recovery reuses exact staged bytes but retains a conflicting temporary file for owner inspection (3360.5267ms) ✔ one Task can seal pending proposals from earlier Runs while snapshots retain each source Run (2704.527ms) ✔ engineering inspection preserves a dirty Git baseline and treats project rules as data (3265.7654ms) ✔ real npm tests and build produce counted results, log artifacts and persistent verification (3272.2014ms) ✔ an actual failing test retains the nonzero exit code and failed count (3101.3505ms) ✔ unchanged automatic checks return original failure evidence, while explicit user reruns and changed source still execute (5820.7951ms) ✔ agent loops cannot hide identical failing checks behind new plan IDs or unrelated reads (13695.5544ms) ✔ zero tests, skipped tests and unrecognized stdout cannot become verification passes (11188.4207ms) ✔ a successful build with no selected tests does not verify the task (3116.6586ms) ✔ stale check plans and forged script names execute no project commands (2146.6904ms) ✔ applying a later proposal invalidates the prior passed engineering snapshot (3161.7513ms) ✔ stopping an actual check terminates its child process and blocks concurrent jobs (2835.0079ms) ✔ check timeout records a failure with its actual termination result (3259.9131ms) ✔ source edits during a successful check make the result stale (4190.522ms) ✔ bounded large logs retain the test footer and readable artifact instead of inventing counts (2799.8124ms) ✔ model-selected checks require a run-bound decision and denial creates no check job (5166.7925ms) ✔ check APIs expose owned plans and jobs and preserve the authoritative results across restart (7871.6252ms) ✔ reviewed changes continue the original task without reopening a generic completed run (4310.8769ms) ✔ concurrent review-continuation API requests accept one run and retain the goal after restart (11501.2097ms) ✔ the model reads owned failure logs without running the check again (8304.4713ms) ✔ process exit at check queued preserves results and never replays commands (5319.4896ms) ✔ process exit at check result_committed preserves results and never replays commands (1881.4826ms) ✔ model code tools keep Task/Run ownership, index evidence and exact proposals across storage recovery (8028.6037ms) ✔ bundled code APIs isolate project results, validate queries and create review-only patches with conflicts (3518.3576ms) ✔ code tool arguments reject malformed ownership-independent inputs before side effects (0.3918ms) ✔ observe-only device authority cannot read, configure or clear project parsing caches (1593.2139ms) ✔ pairing is short lived, single-use, generation-bound and validates exact permissions (1823.2913ms) ✔ device transaction failure neither consumes the ticket nor leaves a credential (1094.4081ms) ✔ devices have project and action boundaries; rotation, expiration, revocation and host changes reject old authority (1640.3441ms) ✔ persistent registry stores hashes and a bounded audit, never raw credentials (1317.656ms) ✔ real TLS verifies IP SAN, CA chain, encrypted restart and public fingerprint (2258.942ms) ✔ TLS sealing failure leaves listener closed and no plaintext or successful TLS registration (865.0898ms) ✔ HTTP pairing is concurrent single-use; secure cookie and strict origin/host checks (985.1782ms) ✔ gateway never exposes management routes, arbitrary assets or unpaired project data (1260.8957ms) ✔ authorized project events paginate from original seq without leaking other project events or payloads (1918.8186ms) ✔ remote task projection excludes raw model history, host configuration, tool input and credentials (968.8877ms) ✔ permission checks apply to direct remote mutation, not only UI controls (1159.0008ms) ✔ approval fingerprint and exact task/revision are bound; duplicate requests never execute twice (1769.807ms) ✔ unknown original operation is recorded and never replayed after response failure (1704.2124ms) ✔ revocation and gateway closure reject old access without invoking runtime stop (1273.2178ms) ✔ rate limits reject pairing brute force and do not persist candidate secrets (1067.7098ms) ✔ real Windows CurrentUser DPAPI persists only a sealed CA and reopens it (2851.7308ms) ✔ mobile approval executes the original Runtime command exactly once and rejects the later desktop decision (3406.2917ms) ✔ desktop rejection wins a concurrent mobile request and original command remains unexecuted (3530.3365ms) ✔ phone stop waits for the original process exit and prevents later heartbeats and command replay (3905.8508ms) ✔ Run replacement during parent-stop wait is rejected rather than stopping the new Run (571.5631ms) ✔ diff pages preserve source SHA; report reads verify blobs and exclude unrelated artifact kinds (1242.0168ms) ✔ hot upgrade refuses an active gateway and restores closed gateway ownership on failure (1216.5799ms) ✔ damaged device permissions or expiry fail closed and never overwrite the original registry (1210.782ms) ✔ device revocation destroys its actual authenticated TLS connection (994.8847ms) ✔ companion observes only current Task screenshots, verifies original PNG bytes and loses access immediately on revoke (1607.6288ms) ✔ mobile browser observation is current Task/project scoped, preserves original PNG and exposes no browser control or profile data (1735.8969ms) ✔ actual config HTTP failure preserves the committed selection, config and volatile credential before a successful retry (1677.8507ms) ✔ controlled installation pins the native kernel and bridge independently of the existing connection-probe identity (1333.648ms) ✔ controlled profiles disable every native project/terminal provider and native model/retry route using the actual include patch contract (0.5662ms) ✔ owned tool catalogues seal full schemas canonically and reject duplicate or malformed declarations (5.0918ms) ✔ bridge authentication, current Run, process, SDK session and browser-origin checks reject before dispatch (867.055ms) ✔ bridge duplicates, unknown tools, malformed arguments and source changes are refused before a second side effect (1327.164ms) ✔ bridge model requests bind the exact catalogue, route and output bound and preserve raw request/result identity (880.1395ms) ✔ stopped and oversized bridge requests never reach a local model or tool callback (655.9799ms) ✔ a Windows virtual data-directory name binds new storage and credentials to its actual physical root (1931.4451ms) ✔ physical data-root adoption still rejects explicit junctions at the directory or any ancestor (919.6246ms) ✔ actual Windows CurrentUser DPAPI roundtrips owned UTF-8 and rejects damaged bytes and wrong entropy (1601.3842ms) ✔ only explicitly supplied keys persist; new provider restores original revisions and public views contain no key (3562.3972ms) ✔ registry and cipher commit failure restores original routing, keys, block and every database row (2899.2756ms) ✔ a two-key import with a protection failure rolls back all registry and cipher entries and retains its preview (3926.4997ms) ✔ connection revision changes retain the original saved block and report a source mismatch on restart (5364.955ms) ✔ disable skips real restart loading, explicit enable restores, delete preserves memory and forget removes both (3053.1027ms) ✔ corrupt saved block and another SID do not load; failures are visible without returning ciphertext or plaintext (3508.7095ms) ✔ legacy persistence requires a new explicit key and binds config and credential revisions; stale requests cannot save (3001.4008ms) ✔ owned plaintext keys never reach SQLite pages, WAL, config or public execution exports (1968.0122ms) ✔ reviewed application invalidates old checks, names current SHA and advances through a new approved check to Git preview (9364.8798ms) ✔ skip-tests, foreign task, unknown operations, diagnostic and parent review retain their boundaries (4651.4296ms) ✔ before the first change, real baseline failures persist across later reads; passed baselines and changed sources cannot certify delivery (9466.1754ms) ✔ actual process exit at Git queued never replays a commit and can reconcile the durable result (15495.0947ms) ✔ actual process exit at Git preparing never replays a commit and can reconcile the durable result (17222.019ms) ✔ actual process exit at Git index_locked never replays a commit and can reconcile the durable result (16987.3936ms) ✔ actual process exit at Git prepared never replays a commit and can reconcile the durable result (20234.4209ms) ✔ actual process exit at Git ref_prepared never replays a commit and can reconcile the durable result (23193.7848ms) ✔ actual process exit at Git publishing never replays a commit and can reconcile the durable result (19876.6576ms) ✔ actual process exit at Git committed never replays a commit and can reconcile the durable result (19699.0846ms) ✔ actual process exit at Git index_published never replays a commit and can reconcile the durable result (17616.6962ms) ✔ actual process exit at Git finished never replays a commit and can reconcile the durable result (21552.3059ms) ✔ an unknown Git result cannot overwrite an externally changed index or delete an unrelated lock (27035.6132ms) ✔ known publication with a later source edit retains the commit and marks the delivery for inspection (15952.0102ms) ✔ CRLF normalization, multiple applied revisions and unrelated special index flags preserve exact bytes and ownership (22712.9724ms) ✔ adding new durable record kinds to an existing hot Store preserves all existing Maps and records (10717.2786ms) ✔ an unrelated owner deletion is fingerprinted and preserved while linked task paths are rejected (15636.5742ms) ✔ actual failed tests become a reviewed local commit and evidence report while preserving unrelated index and files (17738.664ms) ✔ selected preexisting unstaged, staged and untracked owner edits cannot enter a task commit (23027.4037ms) Switched to a new branch 'owner-branch' ✔ source, unrelated stage and branch changes after review invalidate a commit plan without changing HEAD (34775.865ms) Switched to a new branch 'delivery-branch' ✔ a later Git metadata change keeps matching source evidence valid but a real source edit requires checks (16017.8826ms) ✔ a corrupt passed-check log and a pending proposal block Git delivery instead of inventing evidence (9481.9649ms) HEAD is now at 02901ef Sealed delivery fixture baseline ✔ Git filters, merge state, detached HEAD and old task ownership are explicitly rejected (45597.7416ms) ✔ local snapshot commits disable repository hooks and signing and isolate the Git child environment (16677.3745ms) ✔ stopping before publication cancels safely; stopping after publication preserves the actual completed commit (40886.9405ms) ✔ delivery API confirms exactly one owned plan, serves verified reports and preserves results across restart (23871.4712ms) ✔ the model can prepare an owned Git review and evidence report while publishing remains a separate user action (11870.1198ms) ✔ HTTP Git review pairs deferred calls, survives reload and never requests more work until the saved preview is resolved (11883.7455ms) ✔ a reviewed Git preview can be revised or withdrawn without publishing or losing the current Task (18008.1088ms) ✔ actual Electron debug argument ordering retains exactly the compiled entry and rejects extra or repeated actions (1.3427ms) ✔ native bridge refuses other origins, subframes, other windows and unsupported inputs before connecting (878.7737ms) ✔ desktop preview opening requires a currently registered project preview and rejects iframe callers (990.3412ms) ✔ file reveal binds the current project, task, path, UTF-8 bytes and original SHA without executing the file (1912.3954ms) ✔ file changed during asynchronous task recheck is not revealed as the old bytes (2633.7237ms) ✔ desktop browser profile and aliases stay outside project inventory and file access (2479.0399ms) ✔ native picker cancel and origin change after dialog publish no project registration (2371.0598ms) ✔ pending native action is not dispatched twice, and current task review returns its original seq (2522.061ms) ✔ native stop cancellation and changed original host never stop a replacement process (2247.1076ms) ✔ window register rejects host PID, stale installation, extra fields and missing creation identity (0.4479ms) ✔ a rebuilt source archive binds actual generated client assets and retains its original running identity after disk changes (3400.5931ms) ✔ host challenge proof rejects a stale challenge, changed identity and changed signature (0.7725ms) ✔ stop confirmation binds the original instance, creation and loaded runtime (0.1344ms) ✔ state paths reject escaping, linked ancestors, ambiguous hardlinks and damaged JSON (4578.0655ms) ✔ failed registration publishes no initialized memory and can complete after its actual inputs are supplied (5820.3837ms) ✔ host shutdown cancels an un-signalled HTTP model request and refuses later requests until recovered (1.368ms) ✔ host shutdown cancels model-pool acquisition and releases queue ownership without inventing attempts (4460.1133ms) ✔ full candidate validation returns the exact loaded-input host identity and rejects undeclared or altered files (3767.1946ms) ✔ fast per-file scan still rejects a junctioned subdirectory, a hard-linked file and a stream path with the original bytes intact (3901.2622ms) ✔ installation requires separate absolute program/data directories and a candidate in its own versions namespace (2945.0832ms) ✔ inventory rejects a new hardlink on an optional file after its full-manifest scan and accepts the untouched candidate after removal (4159.1942ms) ✔ staging retains the original data and pointer without querying or restarting a host (5819.1061ms) ✔ a post-pointer failure restores every original pointer byte and persistent row; a first-install failure removes only its published pointer (13667.9947ms) ✔ changed active manifest and concurrent pointer changes are rejected without overwriting the unknown record (4824.2815ms) ✔ an unregistered non-HTTP listener blocks preflight even when service-info is absent (2877.8176ms) ✔ uninstall archives its original pointer and preserves every data row/artifact instead of recursively deleting data (7408.3993ms) ✔ preflight rejects an unrelated program root or changed native bootstrap while preserving its original bytes (4973.7509ms) ✔ a changed owned bootstrap follows its candidate and a post-pointer fault restores both original bytes (29293.6929ms) ✔ an external bootstrap change after preflight is preserved and never overwritten as an update or rollback (13199.2221ms) ✔ owned startup retries only read-only observation after a slow health response and attests the same original child (8306.0955ms) ✔ unverified startup expires with its original live child and never treats timeout or an unrelated health PID as proof (6263.7578ms) ✔ diagnostic failures and repeated questions retain the original execution error, task, checks and read-only authority (3037.6781ms) ✔ diagnostic binding survives reopen, rejects foreign Task references and detects original error or artifact tampering (1485.471ms) ✔ a new Task never adopts a previous Task error from session.lastFailure (1191.5111ms) ✔ verified research and explicit no-test scopes do not demand an unrelated test job or claim engineering success (2875.4389ms) ✔ unknown scope and unknown operations remain actionable even after a diagnostic answer or read proof (1860.365ms) ✔ delivery status HTTP is read-only, selection-bound and equal to the saved report after host restart (5053.0128ms) ✔ routine engineering facts remove duplicate success excerpts while complete diagnostics and errors remain available (639.1449ms) ✔ a question about the previous workbench error receives local error context and current change states (1613.03ms) ✔ an unknown model request stops, and explicit resume never replays the approved command (5365.0347ms) ✔ exhausted refused connections preserve tool results, and explicit resume adds no new user turn or replay (3560.2768ms) ✔ stopping during model backoff cancels further requests and clears the running request state (1373.9775ms) ✔ HTTP authentication errors are classified, redacted and never automatically retried (17.0092ms) ✔ temporary HTTP service errors retry and recover with recorded attempt counts (23.2454ms) ✔ invalid and oversized protocol responses fail once with a useful response-format diagnosis (5.4855ms) ✔ HTTP rate limits expose their source and obey the service retry interval before recovery (1013.1642ms) ✔ Windows commands preserve Chinese through a PowerShell pipeline into Node (1011.6542ms) ✔ Windows shell syntax errors stay readable and native command exit codes are retained (2131.6528ms) ✔ review and completed states do not turn a generic continue click into a new model request (3474.1972ms) ✔ old session data receives a truthful historical connection diagnosis without rewriting history (4046.1437ms) ✔ engine HTTP checks bind the detected installation and current task, page original bytes and preserve them across host restart (8501.3239ms) ✔ HTTP stop seals a pending SDK call, rejects automatic replay and keeps failures attached to the original session (9105.1914ms) ✔ sealed r1-off failure fits its original 16K config with authentic latest status and readable source (12.8336ms) ✔ sealed r1-on failure fits its original 16K config with authentic latest status and readable source (9.8247ms) ✔ sealed r2-on failure fits its original 16K config with authentic latest status and readable source (7.3896ms) ✔ sealed r2-off failure fits its original 16K config with authentic latest status and readable source (7.5272ms) ✔ archived native call pairs require exact saved arguments/results rather than model-written summaries (8.022ms) ✔ full system/user requirements and mandatory current failure can refuse rather than be trimmed (12.3863ms) ✔ small native input is byte-preserving; developer instructions remain mandatory after compaction (0.165ms) ✔ older failing checks and pending proposals remain visible after a later large file read (14.5235ms) ✔ bounded result keeps every check status and cuts Unicode only on whole characters (0.437ms) ✔ controlled engine HTTP freezes one child, pages original owned input/report, rejects foreign reads and prevents changed-source reuse (24631.4328ms) ✔ a sealed whole-batch review pauses the owned native session without certifying engine completion or losing the saved host result (20557.1026ms) ✔ a changed review seal still closes the owned native process and cannot claim a verified review handoff (17374.4602ms) ✔ native text and call IDs preserve original result pairing while unsupported content fails closed (0.6023ms) ✔ one engine child freezes real installation identity before creating tasks and rejects a changed source (10652.4469ms) ✔ owned native tools return saved results, use real approvals/checks and account each actual request once across new engine Runs (43341.0793ms) ✔ forged native scope cannot change a parent file and a final completion claim cannot replace worker proof (20009.7583ms) ✔ duplicate native IDs are rejected without replaying a saved tool (17438.8326ms) ✔ native input exceeding the context budget is retained and sends no model request (14162.8465ms) ✔ bounded native sends verify original tool provenance and reject forged views or missing versions (18519.538ms) ✔ stop retains a late reported model usage once and executes no tool from that output (31605.7779ms) ✔ a source change after the original model intent stops before native tools execute (33295.8601ms) ✔ an engine source change while approval is pending invalidates the old permission before any check starts (34139.8094ms) ✔ missing or noncompleted turn end cannot be adopted even with an owned receipt and assistant text (0.5379ms) ✔ closed root with an original live descendant remains unknown and retains its process tree (0.2879ms) ✔ host recovery reserves original pending attempts exactly once and explicit review preserves unknown outcomes (21091.3717ms) ✔ corrupted model reservation cannot reduce recovered budget or permit a new engine request (15011.5263ms) ✔ engine detection seals the actual CLI, SDK, profile and Node and rejects mixed versions without launch (1097.2604ms) ✔ owned engine probe waits for real ready identity, verifies six native responses and saves immutable task evidence (1615.1767ms) ✔ an invalid startup nonce is rejected before initialize and preserves the process close and partial report (1668.6178ms) ✔ SDK identity and session directory mismatches retain raw responses and never become verified (5813.3515ms) ✔ cancel during a pending initialize closes only the owned process without sending an unsupported MCP cancellation (2074.0923ms) ✔ cancel before readiness sends no protocol request and a user retry creates a different process and record (2359.1698ms) ✔ source or scratch changes and unexpected model traffic fail after preserving the actual observed result (5189.9058ms) ✔ unowned idle notifications stay quarantined and server requests receive a rejection without granting operations (3327.5241ms) ✔ engine probe environment excludes parent secrets and NODE_OPTIONS while preserving operating system paths (2175.4854ms) ✔ foreign sessions and altered reports or task bindings cannot reuse verified engine evidence (1668.2999ms) ✔ restart never replays an active engine check or stops a historical PID, and keeps prior task state (1558.8867ms) ✔ model argument validation and runtime activity do not grant external engine prompt or project execution (532.8222ms) ✔ a follow-up question receives the actual failed engine check from its original Task and keeps its report unchanged (2947.4282ms) ✔ unfixed dependency URLs and invalid integrity are reported without exposing URL credentials or allowing a plan (1254.5083ms) ✔ hot upgrade refuses an active dependency operation or evaluation before changing runtime objects (0.5307ms) ✔ read-only environment diagnosis identifies actual runtime, missing dependencies and sealed inputs without installing or running scripts (9024.4194ms) ✔ diagnosis rejects unsupported managers, lock integrity and credential configuration without reading config content (9239.5456ms) ✔ approved real npm ci runs only in owned worktree, skips scripts, retains logs and binds later actual tests (15359.4469ms) ✔ explicitly allowing lifecycle scripts executes sealed root script and preserves the separate preparation result (10007.3975ms) ✔ stale lockfile, tampered plan/config, cross-session identities and mismatched download sources never dispatch npm (9311.4739ms) ✔ registry failure keeps original npm exit/log and performs no automatic fallback or replay (10840.1437ms) ✔ stopping a hanging dependency download records only its owned process and preserves stopped logs (11258.1566ms) ✔ restart retains unknown preparation, blocks writes and checks, and requires explicit read-only reconciliation (8353.1742ms) ✔ HTTP environment flow survives refresh, requires explicit approval and preserves task and artifact ownership (16307.8994ms) ✔ Python environment diagnosis reads declarations and runtime identity without executing project code or installing packages (1768.1703ms) ✔ empty build cache directories keep dependency identity while added executable files invalidate it (1299.8779ms) ✔ dependency snapshots reject external links and incomplete budgets instead of claiming environment equivalence (986.6924ms) ✔ model-requested environment preparation uses the original approval boundary and never treats preparation as passed engineering checks (30992.9479ms) ✔ timeout and offline cache miss retain actual failures and new attempts require newly sealed plans (25860.4828ms) ✔ both evaluation worktrees can approve fixed dependency preparation and different actual dependencies reject candidate execution (57589.7927ms) ✔ evaluation denominator distinguishes attempted unsealed candidates from unexecuted samples (0.3259ms) ✔ evaluation summary retains sealed outcomes and stopped or failed original attempts (0.4433ms) ✔ evaluation contract fixes bounded samples, source paths, checks and aggregate budget (1150.2156ms) ✔ same-source workspaces seal real commands before execution and preserve original staged and unstaged bytes (21151.0192ms) ✔ actual original failures and candidate success are both preserved, then adoption requires parent review, checks and Git (69230.6444ms) ✔ candidate cannot shrink tests and changing original test bytes rejects execution before any model request (26773.288ms) ✔ missing upstream usage remains unknown and does not become a free run or invented cost (72298.5402ms) ✔ missing raw check logs preserve unknown result and original sample denominator (63620.5621ms) ✔ source and model changes never silently replace an existing batch input (4169.9564ms) ✔ model revision change before preparing a fixed sample cannot silently use the replacement connection (4386.8608ms) ✔ HTTP evaluations require host token, original fingerprint and session ownership; worker API cannot bypass evaluation flow (24200.6106ms) ✔ sample queue cannot skip or reset attempted samples, and reports freeze further writes (49733.502ms) ✔ stop cancels only current candidate and preserves partial results instead of replaying a request (41620.3878ms) ✔ restart recovers interrupted evaluations as unknown and keeps completed checks and plan identity (30735.4841ms) ✔ JSON migration survives process exit before commit and retains exact legacy backups (2171.2236ms) ✔ process exit after_state rolls back state, event and outbox together (1750.1942ms) ✔ process exit after_event rolls back state, event and outbox together (1776.2312ms) ✔ crash recovery at accepted retains execution identity and never replays side effects (1983.2775ms) ✔ crash recovery at queued retains execution identity and never replays side effects (2224.8422ms) ✔ crash recovery at approval_wait retains execution identity and never replays side effects (2906.9201ms) ✔ crash recovery at approved retains execution identity and never replays side effects (2920.2994ms) ✔ crash recovery at effect_unrecorded retains execution identity and never replays side effects (2926.0111ms) ✔ crash recovery at result_committed retains execution identity and never replays side effects (3616.4571ms) ✔ crash recovery at before_next_model retains execution identity and never replays side effects (5037.1273ms) ✔ state and event failures publish no success and preserve the last committed memory snapshot (864.8699ms) ✔ tool result artifacts and event cursors retain project ownership across restart (2049.7476ms) ✔ token budgets pause before tools execute and continuation keeps original task and accumulated usage (1382.1922ms) ✔ a time budget interrupts an unresponsive model and persists a resumable run (669.7987ms) ✔ a corrupted large result artifact cannot become a recovered success (692.1547ms) ✔ hard-link aliases cannot expose protected bytes through reads, proposals, inventory or automatic project rules (787.6511ms) ✔ ambiguous project hard links are explicit and ordinary independent file copies remain readable (669.9358ms) ✔ current-file HTTP receipts bind Task/Run and bytes, persist on restart and end without altering original runs or user Git data (5324.7409ms) ✔ file targets exclude referenced inputs and prohibited subsequent modifications (1.451ms) ✔ full actual read and existing receipt complete a no-test file task without fabricating changes, checks or commits (2752.2816ms) ✔ partial pages, partial lines, wrong hashes and failed reads cannot be adopted; contiguous full paging can (1504.9302ms) ✔ user edits and new requirements invalidate receipts, and a second task must explicitly adopt a source in its own project/session (1014.094ms) ✔ adopting an unchanged file never satisfies an explicit modification or a new modification instruction (899.5299ms) ✔ required checks and stale passing flags cannot be bypassed by current file receipts (990.1139ms) ✔ results plan proofs bind current file receipts while changes proofs still demand real applied change sets (825.0646ms) ✔ duplicate receipts cannot create progress indefinitely or reset no-progress reads (3507.9294ms) ✔ manual completion preserves the failed original run and refuses unknown operations (798.7995ms) ✔ completion refuses active work, files changed during inspection and a newly selected task (1024.4981ms) ✔ unrelated files cannot stand in for named output paths, nor can a late apply satisfy newer instructions with an old proposal (1274.4825ms) ✔ delivery and mobile projections distinguish adopted bytes, new changes, no tests and no commit; mutations cannot rewrite receipts (1649.897ms) ✔ proposals preserve project bytes until review; apply backs up the exact original bytes (590.8594ms) ✔ new nested files are created only on apply and unchanged content creates no proposal (641.8257ms) ✔ stale reads and changed baselines are rejected without overwriting external edits (755.081ms) ✔ file paths reject traversal, absolute paths, credentials and outside junctions (745.9744ms) ✔ read and search expose bounded text scopes and reject non UTF-8 / oversized files (540.8361ms) ✔ project-local junctions cannot expose or create protected Git and credential files (701.6043ms) ✔ Windows trailing-dot and trailing-space path aliases cannot bypass protected paths (605.8545ms) ✔ a starter is copied into user data once; restarting from another program version preserves existing projects and editable bytes (1338.5347ms) ✔ startup does not insert a demo or rewrite/reorder an existing user project registry (916.9893ms) ✔ PNG bytes are fully decoded, metadata-free and bounded; JPEG EXIF orientation is applied to the sent PNG (107.3118ms) ✔ forged type, short headers, CRC damage, incomplete chunks, trailing payloads, animation and dimension bombs reject (24.4857ms) ✔ draft image ownership, withdrawal, reload and atomic Task binding retain original SHA and original bytes (1157.0543ms) ✔ a failed session transaction restores draft status and cannot leave a Run consuming the attachment (725.4029ms) ✔ draft limits, duplicate selection, source changes and cross-Task association reject before dispatch (1293.362ms) ✔ demo and legacy paths refuse image dispatch without creating a Task or consuming the draft (694.8488ms) ✔ chat-completions sends one owned image per actual request, pairs original tools and verifies exact sealed bytes (3289.2605ms) ✔ responses sends one owned image per actual request, pairs original tools and verifies exact sealed bytes (3342.9609ms) ✔ mixed pool excludes unknown vision before lease and sends zero HTTP to the incompatible member (3126.8129ms) ✔ vision probe requires correct randomized pixel order and evidence expires when the connection revision changes (2300.6577ms) ✔ HTTP upload/read/withdraw uses session authorization, returns byte SHA and refuses unknown ownership (1636.4891ms) ✔ image accounting excludes Base64 as text but seals every original request byte and reserves image input separately (1.5664ms) ✔ actual model comes from owned consumption, excludes other Tasks and deduplicates the current Run (1.542ms) ✔ mixed routes retain each original model/protocol/outcome and unknown use is never reported as zero (0.1372ms) ✔ zero dispatch, missing ledger, missing model and truncated Run history never borrow configured model identity (0.2109ms) ✔ repeated attempt identities across Runs and contradictory usage stay explicit instead of inflating successes (0.1129ms) ✔ observed route never certifies returned identity and retains mismatches and original source (0.1512ms) ✔ directory picker reads one level, preserves empty folders, skips credential and linked folders (1047.8843ms) ✔ HTML preview serves encoded names and relative modules on an isolated loopback origin; closing revokes it (1085.0182ms) ✔ HTML preview refuses credentials, hidden inventories, traversal, external junctions, hardlinks and unsupported file types (923.9455ms) ✔ runtime shutdown closes local preview ports and pending starts cannot escape shutdown (944.6324ms) ✔ idle file previews expire and release their loopback servers (1125.2932ms) ✔ directory and preview HTTP controls require workbench authentication; project pages cannot read bootstrap (4107.5007ms) ✔ MCP registration stays disabled until an explicit real stdio probe and schema-based tool selection, with actual process cleanup (3620.8238ms) ✔ MCP probe holds the project operation lock until its owned process and report are settled (2693.2302ms) ✔ MCP validates the declared JSON schema subset and rejects unsupported schemas before external dispatch (0.8355ms) ✔ actual model MCP call approves long schema-valid arguments, validates the sent catalog and stores one original result without leaking environment secrets (8496.3463ms) ✔ denying an MCP call never starts another process or executes the external tool (7096.0467ms) ✔ actual handshake drift rejects tools/call even if program entry hash is unchanged (11017.5439ms) ✔ cancelled dispatched external side effect stays unknown, closes its owned process and cannot replay on reopen (10638.6237ms) ✔ changed executable input, foreign project IDs and corrupted original contracts block enable and dispatch (4436.6331ms) ✔ registered MCP HTTP API uses fixed execution contracts and project-owned probe records with no model requests (7563.3965ms) ✔ unknown MCP pauses later tool calls and all project writers; fingerprinted acknowledgement clears only the review gates and explicit resume does not replay (11300.4194ms) ✔ MCP acknowledgement persistence failure restores both job and tool review gates; corrupt evidence cannot be acknowledged (8266.5751ms) ✔ actual host exit after an external effect restores unknown jobs without a fabricated report, allows source-bound review and never replays the call (9094.4782ms) ✔ changed enabled MCP programs can be explicitly disabled without launching or trusting their changed source (4138.747ms) ✔ stdio preserves split UTF-8 frames and final replies drained after process exit (5843.1516ms) ✔ server-initiated model requests are refused instead of inheriting provider privileges (2185.05ms) ✔ invalid or over-budget protocol invalid closes only its owned process and is not retried (1333.0102ms) ✔ invalid or over-budget protocol oversized closes only its owned process and is not retried (851.5899ms) ✔ cancel and timeout do not resend a dispatched request and pre-cancel has no request trace (953.131ms) ✔ startup failure remains undispatched and has no process PID to terminate (707.3679ms) ✔ REA launch filters a fixture credential canary and rejects changed entry identity before a new connection (1220.3129ms) ✔ schema1 connections remain readable without inventing observed capabilities or migrating original state (701.1348ms) ✔ protocol, unsupported capabilities and stream choice persist; client claims cannot forge verified evidence (761.8827ms) ✔ ability evidence belongs to exact connection and credential revision; unknown and unsupported reject required routing (0.318ms) ✔ invalid declarations and inconsistent stream settings reject before registry, revision and credentials mutate (896.0284ms) ✔ model registry keeps original credentials/endpoint intact, rejects stale edits, and forgets only volatile keys on restart (1381.7267ms) ✔ config import previews exact revisions, keeps secrets volatile, syncs only when enabled and preserves manual edits on conflicts (980.5142ms) ✔ real HTTP pool fails over 429 to another model and never repeats completed tools; attempts seal actual request bytes and Run identity (5996.6844ms) ✔ pool rotation and weights select members on new Runs; disabled and missing credentials are excluded without requests (8427.3588ms) ✔ 401 blocks the credential revision; updating the key restores availability without leaking error echoes (2079.3699ms) ✔ queued request cancellation and in-flight disable release only owned leases and do not dispatch another HTTP request (737.2376ms) ✔ compiled model-center HTTP API persists public profiles, blocks forged origins, and local operations never invoke a model (2443.984ms) ✔ model registry disk failures restore routing, pool profiles and credential references without partial updates (767.3563ms) ✔ model selection and saved configuration commit together and failures preserve the original mode and endpoint (823.6085ms) ✔ configuration imports roll back the entire batch and retain one usable preview after validation or commit failure (804.1644ms) ✔ HTTP quota retains native code/type/param/request id, bounded original and allowlisted headers, never repeats that connection (2021.9135ms) ✔ Retry-After above 90 seconds remains exact and a short local budget cannot trigger an early retry (1.8122ms) ✔ Retry-After date, expired date, huge value and invalid header retain explicit semantics (1.2602ms) ✔ unclassified 429 remains unspecified and every bounded actual retry retains its own original error (2101.094ms) ✔ oversized HTTP error saves exactly the bounded prefix and labels its incomplete range (1796.9858ms) ✔ Chat SSE error preserves upstream fields, original failed event, partial tool and reported usage without executing or retrying (2080.2809ms) ✔ Responses failed SSE retains original event/response identity and usage (2045.8544ms) ✔ Responses incomplete SSE retains original event/response identity and usage (2010.1553ms) ✔ unsupported successful Responses output retains original JSON and usage with a local protocol origin (2331.3182ms) ✔ HTTP 200 native error is still an upstream failure rather than a fabricated successful answer (0.8647ms) ✔ JSON-escaped credential echoes are redacted in the original preview and capture as well as parsed fields (0.6329ms) ✔ quota failover uses another independent pool member and never replays a failed stream (2678.5955ms) ✔ native parameters are isolated by pool member, preserve Responses state and bind the original output reservation (0.3381ms) ✔ unsupported native fields and invalid values reject before dispatch instead of silently dropping or overriding constraints (0.3196ms) ✔ native settings for legacy and registered profiles persist without changing credentials or old records and reject stale edits (971.1164ms) ✔ compiled native settings API saves/restores exact parameters and rejects reserved fields without mutating the profile (2181.8466ms) ✔ legacy connection test sends its saved native fields and selected output limit field without changing the configuration (2698.6582ms) ✔ identity distinguishes missing, invalid, mixed, exact and explicitly declared aliases without guessing (1.2854ms) ✔ chat-completions JSON rejects a different model before tool dispatch, retaining source and reported usage (2192.5279ms) ✔ chat-completions SSE rejects a different model before tool dispatch, retaining source and reported usage (3072.4576ms) ✔ responses JSON rejects a different model before tool dispatch, retaining source and reported usage (1688.1774ms) ✔ responses SSE rejects a different model before tool dispatch, retaining source and reported usage (1901.9721ms) ✔ record policy accepts missing identity but keeps it explicit, and strict declared mapping is frozen in the consumption (1.254ms) ✔ mixed streamed identities cannot pass strict mode, and interrupted streams retain observed identifiers without authenticating them (1.2107ms) ✔ identity refusal never fails over to another pool member and raw response model cannot leak known credentials (1910.5025ms) ✔ strict refusal remains visible even when the same original response exhausts the budget (1749.6126ms) ✔ strict identity crosses the controlled engine bridge with its original source and zero tool dispatch (22436.676ms) ✔ frozen direct connection policy and delivery reports preserve response identity without filling historical gaps (3246.9559ms) ✔ saved engine budgets verify all Chat/Responses route variants and reject a mismatched output allowance (779.2344ms) ✔ owned CC Switch SQLite imports the selected TOML Responses wire_api and leaves unknown clients untouched (801.4548ms) ✔ Responses executes one original tool through Runtime, carries encrypted Items, reports raw usage and seals exact wire bytes (2572.5769ms) ✔ mixed-protocol pool keeps both actual variants and changes route on 429 without re-executing completed tools (3131.3318ms) ▶ 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" (1729.5273ms) ✔ Reject {"body":{"id":"resp_owned","status":"failed","output":[{"type":"function_call","id":"fc_failed","cal (1599.5377ms) ✔ Reject {"body":{"id":"resp_owned","status":"completed","output":[{"type":"function_call","id":"fc_bad","cal (1549.3391ms) ✔ Reject {"body":{"id":"resp_owned","status":"completed","output":[{"type":"computer_call","id":"not-supporte (1539.9448ms) ✔ Responses incomplete, failed, missing call identity and unsupported hosted outputs never dispatch a tool (6419.2104ms) ▶ invalid JSON in a nonstream Chat or Responses tool group prevents every tool and pool replay ✔ chat-completions (1601.113ms) ✔ responses (1626.9466ms) ✔ invalid JSON in a nonstream Chat or Responses tool group prevents every tool and pool replay (3228.4942ms) ✔ wire adapter preserves tool IDs but never passes Responses-private Items to Chat or another endpoint (0.1962ms) ✔ connection JSON import accepts explicit Responses protocol and continues to reject unknown protocols without reading client sessions (545.6493ms) ✔ reservation uses actual known usage from the same route, never lowers estimates or rewrites original receipts (1.7911ms) ✔ reservation verification binds original Task/Run consumption and rejects changed usage or foreign provenance (0.6725ms) ✔ known high input calibrates the next request and stops before dispatch while preserving the completed read (1599.0672ms) ✔ one reported overrun stops legacy requests and keeps actual use and excess (1479.2692ms) ✔ one reported overrun stops pool requests and keeps actual use and excess (1702.8327ms) ✔ one reported overrun stops connection requests and keeps actual use and excess (2101.5632ms) ✔ reported usage below the limit prevents an unaffordable legacy retry before another dispatch (1671.0652ms) ✔ reported usage below the limit prevents an unaffordable pool retry before another dispatch (1697.568ms) ✔ native bridge seals calibrated refusal with original consumption and preserves versioned request verification (23943.075ms) ✔ original R v1/v2 stay verifiable while v3 excludes the isolated input outlier and keeps the ledger intact (6.9265ms) ✔ v3 never excludes a repeated high ratio, the latest receipt, or a short history (4.4899ms) ✔ reservation proofs reject reordered latest high receipts, altered history and duplicates while old snapshots survive later consumption (5.4789ms) ✔ input-ratio exclusion preserves independently reported output maxima without rewriting historical v2 (3.4006ms) ✔ capability routing skips text-only members before acquiring a lease and records exact refusal in the original Run (1756.7422ms) ✔ all incompatible members reject with no network request, no lease, preserved Task and explicit non-dispatched ledger (1332.1649ms) ✔ explicit synthetic tool probe verifies only actual matching call, never executes it and loses verification on revision changes (777.2942ms) ✔ plain text probe cannot verify tools; wrong tool nonce and client supplied evidence cannot claim verified support (971.4707ms) ✔ a disabled connection invalidates an in-flight probe even when toggle mutates its original registry object (822.9848ms) ✔ selected connection owns the actual endpoint, credential, wire model and mismatch even when legacy is configured for GPT (4335.7572ms) ✔ budget-only API and demo/real selection preserve routes, reject injected endpoint fields and publish authoritative selection without keys (2845.5875ms) ✔ another session cannot change default model, mode or budgets while a request is active (4675.4207ms) ✔ HTTP download preserves original malformed SSE bytes and artifact SHA instead of decoding invalid UTF8 (3098.4808ms) ✔ terminated SSE adapts empty optional tool arguments, retains the exact wire bytes and executes the validated batch once (3823.3893ms) ✔ reported SSE usage survives rejected required arguments "" without dispatching any tool (2515.3192ms) ✔ reported SSE usage survives rejected required arguments "{\"path\":" without dispatching any tool (2772.3956ms) ✔ reported SSE usage survives rejected required arguments "null" without dispatching any tool (2762.2337ms) ✔ a stream without DONE retains reported usage but cannot adapt or execute even an optional tool (2739.4271ms) ✔ a timeout after reasoning-only fragments creates one unknown receipt and never invents reported tokens (2678.5789ms) ✔ real Chat SSE carries UTF8 and split tool JSON; Runtime executes only after DONE, stores partial progress and original usage (4292.0228ms) ✔ Responses typed SSE validates original item identity, terminal data, raw usage and tool result pair (4235.4ms) ✔ incomplete network stream retains unknown attempt and never automatically switches pool or dispatches fragment tools (4242.3938ms) ✔ stopping an active HTTP stream closes only original request and keeps partial text without executing tools (4186.0521ms) ✔ explicit streaming probe uses actual SSE despite unknown declaration and binds its own original attempt (1518.9498ms) ✔ coalesced initial SSE events publish the latest partial state even when the upstream stays idle (3437.1186ms) ✔ stream timeout closes the original HTTP response, preserves redacted capture and never retries (1413.3896ms) ✔ stream idle timeout preserves incomplete tool arguments, records unknown use and never fails over (3449.4415ms) ✔ SSE accepts CR-only final event boundaries and retains exact invalid UTF8 bytes without dispatch (0.7452ms) ▶ stream parser rejects truncation, finish length, malformed JSON, duplicate sequence, changed final item and oversized data ✔ reject 0 (0.194ms) ✔ reject 1 (0.1688ms) ✔ reject 2 (0.1338ms) ✔ reject 3 (4.918ms) ✔ reject 4 (0.3102ms) ✔ reject 5 (0.4496ms) ✔ stream parser rejects truncation, finish length, malformed JSON, duplicate sequence, changed final item and oversized data (6.7044ms) ✔ compiled host retains interrupted HTTP 200 and explicitly resumes after the recorded cooldown (15216.4995ms) ✔ compiled host retains interrupted HTTP 503 and explicitly resumes after the recorded cooldown (16831.4319ms) ✔ owned IPv4 service works after a failed IPv6 candidate without a second model POST or global DNS change (26.1669ms) ✔ environment HTTP proxy tunnels the owned request and NO_PROXY sends a separate direct request (10.4356ms) ✔ original connect timeout has a bounded nested cause and does not masquerade as response generation timeout (0.7406ms) ✔ POST accepted then reset is unknown, is charged once and never automatically retried (4.4227ms) ✔ partial HTTP bodies retain original request id and bytes, and an unknown pool attempt never switches model (4798.269ms) ✔ response wait timeout after an accepted POST is unknown; cancellation closes the owned socket and permits explicit reuse (148.9585ms) ✔ older configurations receive bounded wait defaults without changing the original object (1.1812ms) ✔ active stream survives sixty seconds; byte activity resets idle but cannot extend total deadline (0.7713ms) ✔ silent stream has its own accurate deadline; caller cancellation and cleanup preserve their cause (1.3724ms) ✔ waiting for headers is bounded independently of stream-idle time (0.2146ms) ✔ wait settings persist, pin new runs and leave model pool, credentials and old run snapshots intact (2375.2684ms) ✔ plain endpoint checks only TCP, records family coverage and never reads model credentials or writes routing/health/attempts (2237.5763ms) ✔ local DNS failure is separate from TCP/TLS and arbitrary input, stale revision and TLS overrides are rejected (2146.8695ms) ✔ refused TCP socket preserves its stage and cannot masquerade as a TLS or API response (2539.7704ms) ✔ TLS validation succeeds for a trusted owned certificate and rejects the same untrusted certificate (2975.2871ms) ✖ a TCP peer that never answers TLS is bounded, and cancellation waits for owned sockets to close (1896.9409ms) ✔ a stalled lookup is cancellable and the overall deadline bounds DNS without late socket work (1535.2252ms) ✔ dual-stack fallback and explicit family results remain separate and never change global DNS defaults (1765.3154ms) ✔ proxy settings are reported as scope only; diagnostics do not contact the proxy or leak its credentials (1591.3232ms) ✔ history is bounded, survives reopening, marks interrupted work and reports changed revisions without rewriting frozen targets (2116.9606ms) ✔ HTTP API, host activity, configuration edits, model start and upgrade honor the same active diagnostic and stop operation (1743.3046ms) ✔ provider shutdown cancels diagnostics and permits explicit reuse after a failed host stop (1872.7248ms) ✔ unchanged proposals cannot hide repeated original-log reads or identical session inspections (13520.5211ms) ✔ reusing a verified proof never resets identical-state detection or creates a new proof (11288.6535ms) ✔ coverage merges adjacent pages, separates changed sources and never claims old Run coverage in a continuation (0.7203ms) ✔ paged inspection finds a corrupted old artifact outside the recent fifty and retains original evidence without model/tool execution (11049.7503ms) ✔ inspection rejects ambiguous/out-of-range pagination and reports unknown tool/task identity without resolving it (2194.1024ms) ✔ original source locks all 60 byte-identical files including modules outside the nine-file prompt experiment (1670.5499ms) ✔ a new original executable or missing license cannot silently join the fixed trusted-source profile (2549.3617ms) ✔ original mode has a separate source identity and does not accept retry or privilege fields (2646.7148ms) ✔ actual original model input requires the complete postPrompt once in system context and cannot borrow a user or tool copy (0.3482ms) ✔ Windows directory permissions normalize overlapping grants without allowing an unrelated sibling (3138.8086ms) ✔ durability broker rejects foreign authentication, outside paths and wrong file identity before real fsync (2677.6507ms) ✔ exact patches preserve BOM/CRLF, unicode and all unselected bytes and back up the original on apply (3543.2788ms) ✔ several patches are matched against one original snapshot in any input order and keep LF (2056.378ms) ✔ mixed line endings require exact anchors and preserve untouched formatting (1638.4871ms) ✔ ambiguous, overlapping, missing and malformed patch anchors never create a proposal or write (1334.4019ms) ✔ stale original hashes and changes before apply preserve owner edits and mark conflicts (2087.2448ms) ✔ no-op BOM patches/full content produce no change and replacements cannot split Unicode bytes (1164.5026ms) ✔ real owned Vite starts only after approval, records OS ownership and live changes, and stops its tree (17721.5249ms) ✔ occupied ports and approval-time conflicts never take over an unrelated listener (14988.0099ms) ✔ background preview final-log failure retains unknown evidence without an unhandled startup promise (19355.1591ms) ✔ stale source, tampered plans, wrong ownership and arbitrary scripts cannot dispatch a preview (14109.0691ms) ✔ restart retains original unknown preview, blocks mutation, reconciles absent identity without replay (16930.4464ms) ✔ an actual shell command keeps Node and UTF-8 output while omitting host-only credentials (1547.3306ms) ✔ real npm tests and build do not inherit arbitrary host credentials or Node injection settings (4267.4016ms) ✔ real Python tests and compilation keep their sealed inputs without host credentials (5743.6566ms) ✔ a process exiting between the Windows snapshot and owner query is absent only after a fresh query (1944.9621ms) ✔ owner observation errors remain errors while a process is present, replaced, or cannot be rechecked (2498.7336ms) ✔ cache controls require desktop authority, exact inputs and owned project identity while preserving task evidence (5217.9004ms) ✔ cache enablement and derived reuse survive a real host restart without becoming task recovery state (7781.4893ms) ✔ persistent parsing is project/SHA/config bound and produces identical semantic indexes after restart (3632.1514ms) 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 (3284.9629ms) Switched to a new branch 'other' ✔ branch and commit changes invalidate even byte-identical sources (7663.2428ms) ✔ corrupt, malformed and schema-invalid records fall back to fresh parsing (1950.3014ms) ✔ disable and clear persist without touching sources, sessions, configuration or original artifacts (2332.5963ms) ✔ cache capacity bounds disk and memory by bytes and LRU project count (3001.5624ms) ✔ incomplete scans do not load or save cached entries; unavailable files leave the current result scope (3239.6621ms) ✔ project and cache junctions cannot expose or delete an outside directory (1598.2733ms) ✔ concurrent file changes refuse a snapshot and clear during pending work cannot repopulate it (1010.6518ms) ✔ file access mode changes invalidate parsed material even when bytes are unchanged (1724.3497ms) ✔ tracked directory replaced by a junction invalidates access scope and never uses its old cache (6750.7696ms) ✔ JS/TS/React AST indexes actual declarations, exports, imports and verified source lines (2286.6896ms) ✔ default expressions, named default exports and declared functions retain their AST evidence (2002.7403ms) ✔ incremental refresh invalidates changed/deleted symbols and recalculates reused import targets (4056.953ms) ✔ Python candidates exclude comments and multiline strings and declare limited evidence (1832.5912ms) ✔ indexes isolate project identity and root and evict bounded caches without sharing results (2227.382ms) ✔ Git ignore, credential paths and links stay outside the index and no project files are executed (4325.2034ms) ✔ malformed, binary, non-UTF8 and bounded files expose partial coverage while good files remain searchable (2065.7105ms) ✔ index query rejects invalid limits/types and does not imply reference or alias resolution (2481.2517ms) ✔ read-only engineering inspection disables configured Git clean filters and preserves index bytes (2092.165ms) ✔ Python multiline declarations retain candidate scope without interpreting expressions as declarations (1103.9554ms) ✔ TypeScript module candidates support import-equals and MTS/CTS while opaque imports stay explicit (1535.411ms) ✔ actual HTTP requests reuse only this project source-checked memory as redacted data and preserve sent evidence after archive/restart (6643.7856ms) ✔ actual HTTP request excludes archived, timed-out and foreign-project memories (6756.1274ms) ✔ actual next HTTP request removes expired memory after a tool observes a source change without replaying that tool (5041.1903ms) ✔ incomplete inventory and excluded-file SHA changes cannot enter actual HTTP memory context or inspect_session snapshots (4782.2261ms) ✔ corrupt original artifact expires its memory and cannot reach an HTTP model request (4465.5112ms) ✔ memory JSON is bounded in escaped UTF-8 bytes and entry count, skips oversized entries and retains complete multilingual values (5959.1201ms) ✔ optional memory yields to the complete goal and original code evidence under the whole-request budget (2232.9808ms) ✔ changed sealed memory artifacts or binding fields fail verification while the original request remains immutable (3352.1692ms) ✔ a malformed project declaration omits optional memory while the repair Task still reaches the owned HTTP service (3929.1793ms) ✔ an expiration persistence failure prevents dispatch and is not hidden as an optional scan failure (2248.5331ms) ✔ project memories persist across restart, keep source provenance, and are visible across tasks (2547.9336ms) ✔ source changes expire current project memories and keep the original evidence (2118.1663ms) ✔ incomplete source scans mark current memories unverified without expiring or reusing them (1678.31ms) ✔ explicit file provenance is rechecked even when the inventory excludes that path (1369.2146ms) ✔ execution snapshots do not expose project memories before their source scope is validated (1739.8817ms) ✔ project memory records stay isolated and archive requires the owning session (2127.5399ms) ✔ project memory HTTP API exposes read-only source status and explicit writes (4022.2253ms) ✔ two Purge source versions lock nine original inputs without executing JavaScript or postinstall and remain disabled by default (991.8789ms) ✔ incomplete, unsupported, relative and oversized sources cannot enable the experiment (714.3799ms) ✔ invalid source fingerprints and invented privilege or retry fields fail before any parent Task is created (4498.5665ms) ✔ source edits between detection and plan are rejected without creating a parent Task (3530.6595ms) ✔ source edits after sealing cannot launch a child and do not reset its cumulative budget (15871.2284ms) ✔ literal experiment text is sealed into the parent, worker and each new Run while actual host approvals and saved results remain required (41230.7864ms) ✔ source changes after a model response stop before any native tool is accepted, retaining the actual usage once (17672.3803ms) ✔ source changes while approval is pending invalidate the old permission before an engineering check starts (17635.8356ms) ✔ altered text, interpolation, section order and policy flags cannot reuse a sealed experiment (1856.1828ms) ✔ the observed SDK system-message layout and the explicit system field both require exactly one effective section, never a user or tool copy (2963.1281ms) ✔ a missing or duplicated experimental section in the actual input is rejected before any model dispatch (25456.0644ms) ✔ experimental prompt overhead still participates in the original complete-input budget guard (14016.2192ms) ✔ source detection HTTP remains local-authenticated and read-only with no configured model requests (2541.5397ms) ✔ Python HTTP plans keep npm defaults, require owned approval and preserve actual results across restart (16790.3128ms) ✔ an isolated three-module Python Task repairs, checks and delivers a source-bound Git report without extra model calls (54611.5373ms) ✔ Python static inspection does not import project code or probe a project interpreter before approval (1500.2454ms) ✔ a real three-module Python failure, reviewed repairs, tests and compilation retain exact structured evidence (5120.1309ms) ✔ unittest counts repeated execution ordinals and subtest skips rather than deduplicating display names (3896.4781ms) ✔ unittest empty, skipped, expected failure and class initialization errors never verify a task (9559.7714ms) ✔ stdout claims cannot replace actual unittest results and an early exit cannot forge a successful collector (3853.1645ms) ✔ compilation reads actual Python source bytes without executing top-level project code (4289.3889ms) ✔ Node and Python mixed projects retain the existing npm default and require an explicit Python selection (3875.8531ms) ✔ changed project interpreters and forged Python plan bindings execute no check (2624.5875ms) ✔ malformed TOML and invalid Python test paths are reported without expanding npm check scope (1619.8239ms) ✔ stopping a real Python check terminates its child and keeps the completed result as cancelled (2794.8434ms) ✔ real pytest pass, failure, skip, collection errors and teardown errors use collected case outcomes (21923.4639ms) ✔ a real project virtual environment is detected statically and its actual version is recorded only during the approved check (10412.1243ms) ✔ a missing interpreter and repeated pytest item objects cannot create a false verification pass (4129.4907ms) ✔ a mixed turn with new bytes and three repeated modules advances, while repeat-only and independent no-op guards still pause (0.9738ms) ✔ runtime sends the next real HTTP request after a W-shaped mixed-progress turn instead of stopping on its trailing repeats (7940.5988ms) ✔ continuation reads advance through historical pages without a false pause while repeated current-Run pages still pause (1.2509ms) ✔ a new tool page replaces old latest-feedback space and still obeys its escaped protocol budget (0.382ms) ✔ budgeted artifact pages count escaped protocol bytes and reconstruct original multilingual JSON without losing characters (9.2867ms) ✔ known artifact coverage detects semantic repeats despite changing limits, and advancing pages or direct evidence restore progress (0.3839ms) ✔ actual large artifact reads are present in the next bounded HTTP request and next_offset progresses without redoing tools (10096.1403ms) ✔ a real repeated artifact loop pauses the same Task after recorded results, survives reload and only resumes explicitly (12202.9064ms) ✔ actual HTTP continuation after reload sends distinct historically covered pages without false no_progress or replay (10405.886ms) ✔ continuations distinguish unread source, later original line ranges and already returned result wrappers (1891.5416ms) ✔ pure-file batch returns unused small-page space while mixed and historical allocation stays conservative (0.3217ms) ✔ v3 pools mixed file/log reads, serves small sources first and leaves historical/mutating batches unchanged (0.2783ms) ✔ v3 W-shaped mixed reading returns complete small modules and substantially larger first test pages in the next bounded request (7531.8373ms) ✔ observed upstream high input pauses before a new request, preserves reported usage and distinguishes the limiting ledger (823.3054ms) ✔ chat-completions: complete actual Unicode/CRLF pages and small source survive the next bounded wire request (7605.6841ms) ✔ responses: complete actual Unicode/CRLF pages and small source survive the next bounded wire request (5755.9333ms) ✔ anthropic-messages: complete actual Unicode/CRLF pages and small source survive the next bounded wire request (5219.5104ms) ✔ an installed app outside the source drive resolves registered SDK and Purge through actual input verification (953.2724ms) ✔ standalone registration resolves the upstream metadata directory and missing upstream falls back to the other registered source (1009.5534ms) ✔ registered hints keep explicit roots, environment and existing app-relative candidates in their original priority (1210.5887ms) ✔ unknown IDs, duplicate registered IDs, wrong versions and relative or inconsistent metadata are not lookup candidates (1171.2802ms) ✔ oversized, malformed, invalid UTF-8 and non-file registry inputs cannot silently enable a source (786.4229ms) ✔ a linked registry directory is rejected without reading or deleting its external source (580.5688ms) ✔ matching research declarations do not replace actual installed version and critical-source checks (723.0145ms) ✔ runtime guides map archived evidence while all original reports and successful/failed source bytes remain independently verifiable (988.3169ms) ✔ missing required research reports fail packaging rather than creating a successful placeholder (882.2138ms) ✔ full verification reports only verified bytes, complete declared scope and actual directory inventory; changed/deleted files still fail (1100.1285ms) ✔ preflight and activation continue one monotonic operation bound to installer hash; terminal replay and changed origin are rejected (8907.1716ms) ✔ preflight failures and fixed-candidate mismatches retain the actual reason without publishing a pointer (5877.4204ms) ✔ cancel finishes only the original ready preparation, preserves data and pointer and refuses replay (7862.7096ms) ✔ the installation lock rejects concurrent same-id operations before they can overwrite original progress (4832.1673ms) ✔ original post-publication failure is linked to its rollback receipt and every original byte survives (6653.7719ms) ✔ a candidate mutation at the publication boundary still fails even after an earlier successful full validation (9714.0438ms) ✔ public progress marks a lost process unknown without editing its original record or claiming completion (6206.5611ms) ✔ throwing console/UI progress observers cannot decide the installation result (7891.0723ms) ✔ a progress-only write failure after committed publication returns the authoritative receipt and a warning (4271.1913ms) ✔ browser and MCP approvals use actual tool categories; mobile exposes only category while keeping original fingerprint (965.797ms) ✔ the compiled runtime version constant tracks package.json so a release bump cannot ship a stale served version (10.4388ms) ✔ context APIs retain exact request identity, paginate archived groups and reject foreign sessions/runs (3783.1008ms) ✔ delivery reports verify request body, sealed plan and original archived bytes, and show corruption as needing verification (2849.3053ms) ✔ native seed framing reserves capacity before optional history without changing the selected window (1069.1793ms) ✔ direct code evidence remains selected while recent historical-index browsing is archived within the complete request budget (2850.8932ms) ✔ one long user turn becomes bounded complete groups with exact UTF-8 history and recent JSON arguments (5837.6879ms) ✔ tool definitions, exact system bytes, output reserve and remaining Run/Task budget are accounted before requests (810.4583ms) ✔ oversized required user instructions pause intact and can be explicitly resumed after increasing the window (1907.8386ms) ✔ all task instructions, constraints and acceptance remain in the runtime system instead of an eight-item or byte prefix (2101.8304ms) ✔ a newest manageable multi-tool group stays exact even beyond the working target; oversized earlier groups are archived (897.589ms) ✔ invalid orphan result is rejected before archive or network preparation (679.3146ms) ✔ invalid missing result is rejected before archive or network preparation (798.5945ms) ✔ invalid wrong result is rejected before archive or network preparation (1135.9733ms) ✔ invalid duplicate call is rejected before archive or network preparation (879.6104ms) ✔ a group larger than the artifact limit uses verified Unicode-safe chunks without replacing tool JSON (1915.8082ms) ✔ changed or missing archived evidence stops reuse and is never silently recreated (1684.3471ms) ✔ transaction failure prevents publishing a plan and any corresponding model request (1451.5286ms) ✔ the Provider rejects body mutation after a sealed plan before calling HTTP (1474.636ms) ✔ HTTP failure and explicit same-Task resume reuse exact completed results and request hashes without repeating a command (11718.1459ms) ✔ a real process exit after plan persistence marks interruption and never issues an automatic model request (1743.8139ms) ✔ local draft preserves exact original ranges and duplicate source instances without inferring outputs or accepting facts (859.8498ms) ✔ review keeps unknown blocking, references distinct from output and boundaries distinct from compliance evidence (2353.2613ms) ✔ draft adoption must account for every original range and does not fill paths from a reference (1855.3828ms) ✔ source and requirement changes, other Tasks and tampered draft bytes are rejected without losing originals (3617.9817ms) ✔ partial explicit source review leaves the other source unreviewed while old file evidence remains valid (1406.5174ms) ✔ bounded draft does not claim full source coverage and cannot acknowledge an omitted remainder (654.776ms) ✔ draft HTTP preserves its original artifact across restart and rejects stale scope and source versions (5021.7538ms) ✔ requirements HTTP binds selection, enforces optimistic revisions, restores original proofs and leaves provider history untouched (5225.0615ms) ✔ research evidence binds actual original bytes; model completion is not user acceptance (1754.6945ms) ✔ same text with distinct user source instances stays distinct; an appended requirement only invalidates its own revision (1645.326ms) ✔ foreign project, old Task and fabricated evidence strings are rejected with preserved failed proofs (4734.0011ms) ✔ source changes invalidate prior proofs and cannot certify a read that was already stale at binding (4545.9056ms) ✔ missing or tampered artifacts never become evidence and original failures remain after recovery (3325.6357ms) ✔ checks omitted by explicit agreement are reported as not executed and never passed (1501.4365ms) ✔ user confirmation stays separate from engineering evidence, is bounded to current inputs and cannot be silently converted (1775.0524ms) ✔ existing multi-file results keep provenance and invalidate only evidence whose files changed (2745.2533ms) ✔ unrelated additions preserve the checked file facts but remain visible as unreviewed user sources (2275.0992ms) ✔ actual Git preview and publication recheck changed requirement bindings without touching HEAD or the owner index (11576.6488ms) ✔ no-test file application and manual completion keep pending user requirements for explicit review (5608.4254ms) ✔ revision races, forged source ranges, evidence fields and deletion cannot overwrite sealed requirements (1423.8327ms) ✔ requirement binding prevents completion and Git acceptance after a review changes; legacy tasks stay explicit (4694.2939ms) ✔ deep Windows data directories keep REA output within the external path allowance and retain owned evidence (4215.4469ms) ✔ user research registration creates an owned Task without model requests and seals the exact artifact identity (2431.7782ms) ✔ complete research binds original Evidence and Bundle, pages exact fields and reuses verified results after reload (4299.2135ms) ✔ artifact changes before analysis refuse dispatch and changes during analysis preserve unverified original responses (4609.9084ms) ✔ owned cancellation closes the connection, concurrent and automatic retries are refused, explicit new analysis preserves history (3060.1814ms) ✔ a dispatched transport timeout stays unknown and reload never launches or kills a saved PID (2658.2215ms) ✔ corrupt Evidence parts and changed binding records are refused instead of silently re-analysed (3434.0363ms) ✔ research documents over the single-artifact limit preserve exact Unicode bytes with owned verifiable parts (1713.2876ms) ✔ research tools are only sent for registered Task artifacts and actual HTTP uses the sealed tool contract without replay (16297.5742ms) ✔ file identity rejects directory and invalid formats without running a target (1179.5625ms) ✔ a stop from the event loop interrupts immediate model replies before later tools (2855.3288ms) ✔ a stop after the first tool preserves a complete interrupted multi-call protocol group (3062.748ms) ✔ the compatible HTTP loop preserves call IDs, validates errors and waits before executing commands (9426.6608ms) ✔ denial prevents command execution and produces a paired denied result (7098.2124ms) ✔ stopping during approval interrupts pending calls without executing or replaying them (6386.2453ms) ✔ stopping a running fixture command terminates its own process tree and retains uncertain state (6417.833ms) ✔ command output decodes split UTF-8 and excludes the workbench key from child environment (2796.0954ms) ✔ model-list testing uses a tentative key without changing the saved connection and redacts service errors (6.7825ms) ✔ truncated or repeated model calls fail before execution; tool input validation rejects unknown capabilities (11133.4221ms) ✔ request projection keeps exact tool pairs and complete recent user instructions while preserving disk history (3205.9161ms) ✔ local skill revisions require explicit enable, retain original text, survive restart and stay within the selected project (1997.0001ms) ✔ registration rejects non-SKILL paths, oversized/non-UTF8 files, duplicate sources and wrong field shapes without running project code (676.1302ms) ✔ actual HTTP models receive only enabled current-project metadata, then a redacted owned skill page with complete tool pairing (4782.5682ms) ✔ source edits after a catalog request prevent the read and appear as changed on the next request without replay (4541.5743ms) ✔ disabled and foreign skills cannot be read by guessing their IDs from another project (5290.2892ms) ✔ skill catalog takes remaining escaped JSON budget and cannot evict the user goal or current code (4927.0331ms) ✔ multilingual skill body pages fit actual tool JSON bytes and every page uses the same immutable artifact (13021.7247ms) ✔ registry failures roll back enable and refresh state; corrupted snapshots block requests before network (4817.3676ms) ✔ saved skill catalogs retain SHA-bound historical evidence and reject corrupt catalog blobs (7694.9185ms) ✔ HTTP skill administration enforces project scope, revision checks and idle mutation without model calls (6524.9544ms) ✔ JSON state keeps an atomic backup and preserves corrupt primary files (2418.7151ms) ✔ tool-pair recovery fills missing results and drops orphan or duplicate results (0.362ms) ✔ restart pauses unfinished work, retains proposals and never replays a command (2879.4394ms) ✔ public snapshots carry increasing revisions without the private provider transcript or config key (3038.7652ms) ✔ a checkpoint seals all task requirements and original results without a model request and explicit continuation keeps the same Task without replay (11966.6676ms) ✔ checkpoint sources, requirements, plan and original tool changes block continuation and leave original checkpoint bytes intact (3787.9096ms) ✔ unknown tool effects and pending proposals stay blocked in recovery packages rather than being changed into completed work (3564.8139ms) ✔ checkpoint and referenced result corruption are rejected before dispatch and historical blobs are never rebuilt (3355.5881ms) ✔ checkpoint ownership, exact confirmation and concurrent operations reject foreign or stale resumes (4184.8385ms) ✔ Git identity changes and altered plan proofs cannot be accepted as the sealed recovery state (4260.7464ms) ✔ checkpoint records survive SQLite reopen and do not trigger automatic model or tool replay (3240.7365ms) ✔ incomplete scans and in-progress project activity refuse checkpoint creation while private credentials are redacted from saved requirements (4065.6803ms) ✔ oversized recovery evidence is refused before reading or publishing the checkpoint (3941.4461ms) ✔ compiled checkpoint HTTP APIs bind current Task/Run, isolate sessions and only resume on an explicit matching confirmation (9343.8339ms) ✔ actual process exit before checkpoint commit retains only durable records without replay (5467.0528ms) ✔ actual process exit after checkpoint commit retains only durable records without replay (3198.0712ms) ✔ a real two-reply failure replay reaches request three at 16K with complete goal, original sources and sealed tool evidence (12179.5968ms) ✔ request deduplication retains long original intent, extra requirements, constraints and acceptance without changing the stored contract (2338.0065ms) ✔ only the routine context avoids plan echoes; diagnostic API and requests without an included plan keep original observations (756.7482ms) ✔ operational questions are diagnostic while explicit repair requests remain execution instructions (1.0327ms) ✔ an explicit prohibition on running engineering checks is not interpreted as a requirement to run them (1.6901ms) ✔ file delivery requires a positive instruction, not saved evidence or a prohibited write (0.2781ms) ✔ an unsubstantiated text claim cannot complete a requested file delivery (1781.5857ms) ✔ HTML without tests ends at reviewed file delivery and never offers automatic rechecking (3332.2333ms) ✔ diagnostic answers retain the unfinished task, limits and check result, and enforce read-only tools (4196.1251ms) ✔ review continuation is a system action and adds no testing demand or fake user instruction (2895.0686ms) ✔ a provider cannot execute tests against an explicit no-test task even if it ignores advertised tools (2527.6806ms) ✔ a budget pause becomes visible only after its input snapshot can be sealed with the original Run (2071.2139ms) ✔ alternating duplicate file reads and listing files still pauses without treating new pages as duplicates (21275.25ms) ✔ task-plan HTTP mutations bind current Task/Run and retain original evidence through real host restart (5760.7364ms) ✔ concurrent stale HTTP plan revisions cannot replace conditions or attach proofs to a different plan (3906.0056ms) ✔ a real escaped long-file tool message keeps its proof identity within a tight request budget (1393.2209ms) ✔ task plans reject fabricated completion, invalid scopes and cyclic or missing dependencies (0.5344ms) ✔ sealed plans retain goal, contract hash and original steps through SQLite reopen without executing tools (686.2746ms) ✔ proofs require original owned reads, honor dependencies and reject altered invocation arguments and bytes (1510.4229ms) ✔ new user requirements invalidate current proofs and reseal the same original operations in a new revision (1449.015ms) ✔ applied change-set evidence binds reviewed current files and invalidates dependent steps after outside edits (2727.388ms) ✔ actual npm checks are verified by original counted logs and current source, with failure history retained (11984.9957ms) ✔ real desktop and mobile browser proofs bind required actions and PNGs, rejecting source changes, a later failed plan and changed builds (14725.9436ms) ✔ Git delivery blocks incomplete plans and changed preview bindings, preserving HEAD and owner index before a fresh verified commit (21474.6141ms) ✔ the runtime supplies owned read IDs and completes a plan only after original step evidence verification (3043.8499ms) ✔ premature model completion pauses the original Task and resume preserves the existing plan without replay (4385.2135ms) ✔ plan operations participate in project activity and orderly shutdown (1057.0896ms) ✔ actual process exit before plan save restores only committed plan evidence without replay (1332.6705ms) ✔ actual process exit after plan save restores only committed plan evidence without replay (1905.1855ms) ✔ actual process exit before proof save restores only committed plan evidence without replay (1425.0772ms) ✔ actual process exit after proof save restores only committed plan evidence without replay (1298.7105ms) ✔ the latest oversized session result reaches the next request together with already read source modules (3106.7007ms) ✔ every call in a large latest group keeps its actual outcome, including denied and unknown results (2827.2455ms) ✔ a small complete latest tool group is sent once and needs no duplicate feedback view (2049.269ms) ✔ real HTTP repeated state inspection pauses after three identical engineering outcomes and preserves tool pairing (5005.3578ms) ✔ fixed normal exploration, pagination, SHA changes and repeated ranges are external observations, not a score (1.4193ms) ✔ read-only research is valid work; model summaries and unchanged timestamped listings do not become engineering evidence (0.4089ms) ✔ new permissions, network and broken tools are failure evidence; repeated invalid inputs remain candidates (0.4029ms) ✔ waiting for original long checks, review and approvals stays waiting across repeated observations (0.527ms) ✔ baseline facts, real repairs, current-source check results and requirement evidence have distinct growth (0.2135ms) ✔ bounded memory keeps unknown instead of claiming an unretained range is repeated; Task/Run instances cannot mix (12.4683ms) ✔ actual runtime records growing pages then a repeated page without changing model rounds, stop rules or original tool pairs (5679.7906ms) ✔ actual invalid tool calls repeat under the original step budget; observation never adds an early stop (5393.3118ms) ✔ actual command approval is observed as waiting and denial stays a permission failure (5209.9526ms) ✔ network errors remain model failures with shadow failure evidence, never a fabricated no-progress cause (6087.1405ms) ✔ observer receipt failure cannot change tool results, model budget or completion (4900.6358ms) ✔ a later receipt recovers after an observation write failure without misaligning retained source receipts (4961.5963ms) ✔ real applied files, counted checks and current requirement proofs grow after an existing-facts baseline (5894.9459ms) ✔ receipt retention stays aligned at the cap and a foreign step cannot enter the Run (2756.0968ms) ✔ an actual approved long check records its owned running job as waiting before the original counted result (7146.5986ms) ✔ a live SQLite host with the actual 1.9 artifact object upgrades, keeps its original blobs and verifies real browser PNG bytes (7015.3507ms) ✔ failed upgrade restores the original artifact object, provider and request handler (1849.9865ms) ✔ claiming the same runtime owner twice preserves its original metadata and every database row (2620.1193ms) ✔ sessions without controlled-engine history retain their prior export shape and never expose another session record (1524.1151ms) ✔ live workers, controlled drivers, or MCP operations refuse an upgrade before changing handlers, managers or database rows (1340.6516ms) ✔ an idle upgrade loads the newly introduced engine record map, resets obsolete managers and preserves original rows, sessions and provider objects (1511.0357ms) ✔ an installation failure restores old managers, original record maps, handlers and all existing database rows (1674.7503ms) ✔ the owned controlled driver remains an active session and project even after outer task activity has ended (619.9124ms) ✔ release and upgrade help or invalid arguments have no file or network setup side effects in an independent directory (928.6172ms) ✔ hot upgrade adopts a retained virtual data root without reopening SQLite or changing rows, Provider or routing (1368.9252ms) ✔ failed adoption restores the original virtual data-root name and original open database (863.2031ms) ✔ hot upgrade refreshes retained model and importer methods while preserving credential and routing objects (826.4066ms) ✔ hot upgrade refuses active probes and restores retained model methods if installation fails (912.0905ms) ✔ the upgrade compatibility gate includes the served current version and the previous formal patch (0.4321ms) ✔ a directly started compiled host upgrades twice with the same PID, sessions, applied bytes and volatile provider (4430.6178ms) ✔ the running 0.1.0 process upgrades in place and retains its project, session, files and volatile model connection (8141.1223ms) ✔ HTTP worker identities require a current sealed plan, reject overrides and close unused plans without touching source (28957.9146ms) ✔ HTTP unknown recovery only accepts the original evidence SHA and preserves changed source for explicit retention (43640.0306ms) ✔ verified isolated outputs form a new parent batch, rerun real checks and commit only reviewed source (114773.2939ms) ✔ overlapping worker paths are explicit conflicts even if each independent task checks successfully (76423.4572ms) ✔ source edits after isolated work preserve the owner file and invalidate collection without proposing a merge (49912.5394ms) ✔ changed child bytes after parent preview block the entire batch before any parent write (61604.0309ms) ✔ corrupt original check logs cannot become collected outputs or a parent review (51238.0761ms) ✔ stopping a parent accounts late responses and executes none of their proposed tools (34055.7169ms) ✔ unreported interrupted requests reserve estimated usage and the total deadline cannot reset on resume (32700.4144ms) ✔ a deadline before worker request dispatch records no invented usage and cannot reset on resume (29766.3381ms) ✔ a response exceeding the cumulative token allocation stops before registering or executing its tools (38146.591ms) ✔ cumulative tool allocation rejects the entire returned batch before any tool side effect (24171.7135ms) ✔ single response 16 tool ceiling rejects the entire returned batch before any tool side effect (19063.6254ms) ✔ consumed request steps remain consumed across explicit worker resumes (21710.4055ms) ✔ a saved completed response with missing usage is conservatively accounted exactly once during recovery (16604.7141ms) ✔ an unexecuted second workspace plan resumes its original identity without recreating the first checkout (22414.0758ms) ✔ a verified interrupted parent proposal resumes sealing without duplicating or replacing its original change (80793.2714ms) ✔ recovery attaches a sealed parent batch saved before its group reference without creating another batch (53503.3775ms) ✔ explicitly retaining an unexecuted parent merge cancels its batch and frees the next parent goal (49584.5658ms) ✔ unknown collaboration blocks ordinary proposals but permits only original batch recovery and rollback (54500.2346ms) ✔ after a restart the batch application lazily reinstates the worker result guard before any parent write (50208.8797ms) ✔ unknown original tool inputs must match before acknowledgement, and reviewed checkpoint IDs preserve the unknown fact (18803.0148ms) ✔ a parent stop observes and waits for a manually approved child check and its real descendant process (17762.869ms) ✔ a real host crash preserves unknown checks, refuses a still living owned PID and never replays a model or tool (30953.8024ms) ✔ a persisted child initialization resumes the original control run after a failure before its worker record (17047.1417ms) ✔ failed request intent persistence never dispatches a worker request or changes source (17601.4155ms) ✔ failed usage persistence rolls back the uncommitted usage and conservatively reserves one unknown request without tools (17555.2636ms) ✔ a child file change after observation prevents acknowledgement even when the tool inputs did not change (18820.663ms) ✔ worker scopes and allocated budgets reject aliases, extra permissions and aggregate overflow (596.4504ms) ✔ two sequential Git workspaces freeze a new parent goal, separate owned tasks and preserve all source bytes (24105.4187ms) ✔ real isolated changes, approvals and npm checks produce original owned outputs, not model completion claims (43195.8391ms) ✔ forged worker tools cannot execute terminal or Git even when absent from the advertised tool list (15710.907ms) ✔ a completion-only model pauses each worker without fabricated changes or checks (19509.1102ms) ✔ restart retains frozen scope and marks interrupted parent and workers unknown without sending requests (16036.9658ms) ✔ all actually returned same-version fragments reach the bounded request, including overlap and Unicode, with independent original provenance (4043.3613ms) ✔ unread gaps and different range origins remain separate; partial sending names the precise missing already-observed range (818.119ms) ✔ same-SHA conflicting overlaps and invalid coordinates fail closed, while a new SHA never inherits old fragments (933.8648ms) ✔ versioned old views keep the original single-page semantics and verification (773.6895ms) ✔ small code and actual failed command stay visible when their original multi-tool group cannot fit (991.6556ms) ✔ mixed group retains six distinct small file observations with exact multilingual text and bounded request bytes (1421.3766ms) ✔ working observations are bound to the current Task and stop treating a pre-application file revision as current evidence (1911.3518ms) ✔ a sealed observation rejects changed or missing original result artifacts instead of recreating evidence (1594.0872ms) ✔ a real passed-check observation is excluded after a source edit invalidates the check, while its historical sent evidence remains valid (5337.8457ms) ✔ budgeted file pages reconstruct UTF-8 text including long single lines and surrogate boundaries (4.5903ms) ✔ actual HTTP multi-tool reads survive an oversized runtime snapshot and the sealed evidence survives reopening (10915.2707ms) ✔ actual HTTP file pagination advances over a long file with each returned page visible in the next request (8799.1948ms) ✔ later mixed history pages cannot displace the actual module sources and complete requirements (6530.9916ms) ✔ a later narrow page of the same file revision cannot evict the wider actual observation (3081.4401ms) ✔ six small source files stay complete across metadata turns even when mandatory evidence exceeds the soft target (3017.9778ms) ✔ historical read-artifact navigation points to the original artifact and next page, never its result wrapper (1074.0928ms) ✔ a 1.6.1 full-provenance sent view remains verifiable after introducing the compact view (1168.0826ms) ✔ actual HTTP requests keep four observed modules after a later large history read and preserve all source bytes (10824.1238ms) ✔ workspace HTTP ownership, reviewed creation and reopen survive restart without changing source files (11825.1905ms) ✔ isolated model Task applies a batch, executes real checks and commits at the Git review boundary with a source-bound report (39083.4702ms) 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 (7622.7201ms) ✔ an ignored external file at checkout is observed and never accepted as pristine creation (6023.106ms) ✔ outside content hidden by assume-unchanged cannot pass workspace verification and blocks further source execution (5281.4561ms) ✔ workspace preparation refuses known shared-repository activity and preserves archived child sessions (6522.5275ms) ✔ compact checkout identities create in deep Unicode source directories while retaining the human branch name (7098.32ms) ✔ Windows Git metadata path limits reject creation before publishing a branch or changing source bytes (3204.9868ms) ✔ a pending source snapshot participates in shutdown and completes before storage can close (2265.7442ms) ✔ source configuration values stay private and a configuration change invalidates a saved review (3140.8416ms) ✔ workspace review rejects excluded dependency before any checkout (2804.8093ms) ✔ workspace review rejects large committed blob before any checkout (3045.7656ms) ✔ workspace review rejects case collision before any checkout (3256.1746ms) ✔ workspace review rejects committed symbolic link before any checkout (3106.3361ms) ✔ workspace review rejects external config include before any checkout (2158.2565ms) ✔ actual process exit at workspace intent preserves source and reconciles without creation replay (6052.3212ms) ✔ actual process exit at workspace git_started preserves source and reconciles without creation replay (6991.9786ms) ✔ actual process exit at workspace git_finished preserves source and reconciles without creation replay (8157.66ms) ✔ actual process exit at workspace verifying preserves source and reconciles without creation replay (8240.7338ms) ✔ actual process exit at workspace registering preserves source and reconciles without creation replay (8585.9506ms) ✔ actual process exit at workspace project_registered preserves source and reconciles without creation replay (10100.9075ms) ✔ actual process exit at workspace session_registered preserves source and reconciles without creation replay (10505.0195ms) ✔ actual process exit at workspace ready preserves source and reconciles without creation replay (7629.8737ms) ✔ outside changes at project registration remain unknown and preserve the already registered child identity (9951.1538ms) ✔ managed worktree preserves source index and dirty files, opens a distinct project and binds Run identity (7146.9418ms) ✔ stale review, branch/path collisions and forged ownership do not run worktree creation (3631.9931ms) ✔ workspace project verification is read-only and rejects physical Git identity changes (5305.1841ms) ℹ tests 899 ℹ suites 0 ℹ pass 898 ℹ fail 1 ℹ cancelled 0 ℹ skipped 0 ℹ todo 0 ℹ duration_ms 2808682.1659 ✖ failing tests: test at tests\network-diagnostics.test.mjs:82:1 ✖ a TCP peer that never answers TLS is bounded, and cancellation waits for owned sockets to close (1896.9409ms) AssertionError [ERR_ASSERTION]: Expected values to be strictly equal: 'tcp' !== 'tls' at TestContext. (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/tests/network-diagnostics.test.mjs:86:49) 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: true, code: 'ERR_ASSERTION', actual: 'tcp', expected: 'tls', operator: 'strictEqual', diff: 'simple' }