
> tablaze@0.1.0 test
> npm run build && node --test --test-concurrency=1 tests/*.test.mjs


> tablaze@0.1.0 build
> tsc -p tsconfig.json

✔ workspace export keeps active selection consistent when the active tab closes during storage capture (1349.428333ms)
✔ PDF metadata stays bound to its captured tab through asynchronous artifact persistence (696.56925ms)
✔ drag revalidates observed targets after asynchronous geometry reads before mouse input (564.151625ms)
✔ upload chooser revalidates its visible trigger after asynchronous local file checks (599.939416ms)
✔ newly discovered scroll containers retain node identity guards when replaced (600.881458ms)
✔ scrollable containers expose ranges and accept guarded horizontal and vertical input (526.7ms)
✔ dragging emits the browser drop event and delivers the real data payload (721.0955ms)
✔ hidden file chooser transfers explicit local bytes without a hidden DOM reference (742.171625ms)
✔ PDF exports real private artifacts with verifiable digest and source (562.416291ms)
✔ workspace roundtrip restores isolated durable storage, owned tabs, and active selection (1966.959208ms)
✔ workspace validation precedes navigation and failed restore rolls back owned sessions (848.781792ms)
✔ per-decision catalogs hide execution metadata and context changes skip queued writes (13.330667ms)
✔ a context change during planning or write-ahead persistence never executes the stale handler (2.79825ms)
✔ trusted write effects retain ambiguity for thrown, malformed, error, and unknown outcomes (8.251583ms)
✔ finish rechecks its lease and cannot reuse verification from the old context (1.437584ms)
✔ a context change between closed and newly prepared leases invalidates old verification (1.342291ms)
✔ only a confirmed session close preserves current evidence across context changes and later writes still invalidate it (5.876417ms)
✔ bound checkpoints reject missing runtimes and changed registry or same-origin tenant identities (3.02925ms)
✔ version-one and version-two checkpoints explicitly migrate unbound and cannot gain a registry (1.41825ms)
✔ resume rechecks a pending call effect rather than trusting a tampered mutating flag (1.204333ms)
✔ initialization uses the bound dispatcher and each lease closes before terminal persistence (1.298958ms)
✔ a prepared lease that arrives after cancellation is still closed without planning (3.072791ms)
✔ catalog release failure or cancellation cannot turn accepted verification into success (1.839791ms)
✔ run CLI completes a scripted model task through real Chrome and prints a bounded report (908.560791ms)
✔ run CLI returns human input with exit 2 and an explicit failure with exit 1 (433.919625ms)
✔ run CLI does not accept an unverified model completion and enforces the step limit (234.685ms)
✔ CLI rejects agent-only options outside run and validates required arguments and limits (2406.789792ms)
✔ CLI reports cleanup failure separately from an already verified successful agent result (818.61825ms)
✔ restoration abort handles a rejected asynchronous dispose without an unhandled rejection (336.848708ms)
✔ CLI publishes bounded planner diagnostics without exposing endpoint, key, or upstream error text (237.824875ms)
✔ HTTP planner failures expose only status and fixed classification without enabling default retries (22.675125ms)
✔ unknown exceptions cannot inject diagnostic codes and custom retry policy remains authoritative (3.628625ms)
✔ transport and malformed response diagnostics remain distinct and omit raw response data (3.704042ms)
✔ catalog failure is distinguishable before any planner or mutation runs (1.958459ms)
✔ application callbacks are classified without recursively notifying a broken event hook (6.01475ms)
✔ adapter usage callback failures retain their application phase and original retry policy input (1.228667ms)
✔ persistence failures prevent mutation or successful completion without changing resume schema (1.814625ms)
✔ a noncooperating checkpoint callback receives a bounded persistence-timeout diagnosis (5002.872375ms)
✔ explicit initialization and popup following complete a real workflow without executing remaining origin calls (1244.263084ms)
✔ cancellation during terminal persistence cannot return succeeded with accepted evidence (5.7935ms)
✔ deadline expiry during terminal persistence is reported and returned elapsed time includes the wait (152.958208ms)
✔ returned checkpoint accounts for final persistence time before a later resume (42.928542ms)
✔ resume retains application restrictions unless the caller explicitly replaces them (1.874625ms)
✔ resume cannot silently remove a previously required application completion policy (2.58775ms)
✔ decision checkpoint resume preserves counters and IDs but requires fresh verification (16.343208ms)
✔ an interrupted maybe-applied mutation blocks every resumed planner until trusted reconciliation (3.148333ms)
✔ failed checkpoint persistence prevents dispatch and never retries a tool (2.168583ms)
✔ restored sessions reject old session IDs and refs until a fresh snapshot is obtained (2.282292ms)
✔ untrusted checkpoint shape, protocol pairing, IDs, and budgets are validated before execution (2.407166ms)
✔ stall detection uses repeated actual outcomes, provides one replan, then stops (2.4435ms)
✔ bounded retry and fallback only repeat planning; tools with uncertain effects are not retried (5.086167ms)
✔ opt-in compaction preserves complete tool groups and real verification evidence (8.649917ms)
✔ HTTP adapter reports only actual provider usage and exposes transient status classification (10.824458ms)
✔ separate CLI processes restore private storage and fresh sessions without replaying completed writes (1826.27025ms)
✔ interrupted external side effect requires explicit reconciliation before CLI restores or plans (1710.586084ms)
✔ an exhausted saved CLI budget performs no page restoration or model request (1204.915ms)
✔ explicit initialization traverses actual MCP before planning, with step-zero write-ahead and real evidence (13.693333ms)
✔ start URL is explicit and invalid targets fail before catalog, tools, or planning (0.714292ms)
✔ initialization obeys tool, history, deadline, and cancellation budgets without planning (31.705583ms)
✔ unknown initial effects require reconciliation and the initializer is never replayed after it (1.037959ms)
✔ an initializer that was never registered can start once when its saved prerequisite is supplied (0.618ms)
✔ settled initialization survives resume and compaction with unique subsequent IDs (1.895708ms)
✔ version-one checkpoints migrate explicitly and initialization manifests reject inconsistent history (0.834208ms)
✔ write-ahead persistence failure never dispatches or replays the registered initializer (0.392791ms)
✔ replan_required skips later tool calls and still requires separate verification (0.335875ms)
✔ CLI start URL reaches real Chrome before its first model request and resumes without another initializer (1967.535167ms)
✔ CLI accepts a browser-wide popup policy but restricts start URL to run and validates both (1188.532583ms)
✔ agent drives real MCP browser tools and requires verified form state before finishing (700.249ms)
✔ MCP schema rejects malformed arguments before side effects, then planner can recover (2.646458ms)
✔ stale-reference result stops the batch and is returned for replanning without retry (1.486833ms)
✔ failed checks and isError=true cannot satisfy completion even when passed=true (2.313167ms)
✔ completion rejects stale evidence, another session, invented IDs, and contradicted checks (5.069083ms)
✔ application task-specific acceptance can reject a technically passing unrelated check (1.016708ms)
✔ tool and planning budgets are bounded; invalid decisions never dispatch a tool (1.353875ms)
✔ deadline bounds a non-cooperating planner and abort propagates into an in-flight tool (22.573542ms)
✔ MCP cancellation reaches the actual server request signal (1.132292ms)
✔ HTTP planner preserves MCP image evidence and structured errors without real model calls (10.811333ms)
✔ HTTP planner rejects malformed model arguments, mixed finish calls, and oversized output (0.691875ms)
✔ an installed completion policy must explicitly return true (2.338916ms)
✔ history budget stops later mutations within the same planning decision (0.583292ms)
✔ cancellation before the dispatch microtask never starts a custom tool (0.253833ms)
✔ binding is frozen trusted metadata; observations and DOM edits do not consume it or hold the queue (537.001292ms)
✔ same-URL reload invalidates a document lease even if the cached navigation counter is delayed (565.059792ms)
✔ history routes and tab switch ABA invalidate bindings permanently (632.168291ms)
✔ a navigation during the asynchronous assertion cannot produce a valid stale result (485.433959ms)
✔ signal, explicit close, session close and engine disposal revoke leases without exposing browser handles (718.703209ms)
✔ cancelled acquisition releases a real late Document handle without affecting the session (558.650917ms)
✔ pre-cancelled acquisition creates no remote handle and opaque origins are refused (523.590542ms)
✔ engine disposal revokes an in-flight acquisition and releases its late handle (556.86825ms)
✔ independent CDP navigation invalidates a lease and lease cleanup leaves the external browser alive (611.380541ms)
✔ hotel form batch navigates and independent checks verify real URL and DOM (874.392208ms)
✔ snapshots obey budgets and do not include password or hidden values (364.303667ms)
✔ a newer snapshot rejects the older revision before input (166.315458ms)
✔ a replaced DOM node with an identical label is rejected (175.846083ms)
✔ changed target semantics are rejected before a click (169.624583ms)
✔ overlay blocks input and retry requires fresh observation (1252.83425ms)
✔ semantic change during actionability waiting prevents the delayed click (434.019458ms)
✔ a base URL change invalidates the observed resolved link destination (236.066375ms)
✔ value verification never leaks a password substituted after a field-type read (249.299625ms)
✔ batch preserves completed work and reports failed and skipped steps (235.976041ms)
✔ incremental snapshots expose added and removed controls (308.744458ms)
✔ open shadow controls and explicitly selected frame controls execute (343.748416ms)
✔ sessions isolate storage and reject cross-session references (408.96975ms)
✔ concurrent operations serialize within one session (238.997458ms)
✔ capture contains a real JPEG and closing removes a session (193.051459ms)
✔ CDP disposal closes only owned pages and preserves the independent browser and existing state (904.447709ms)
✔ stream and item error diagnostics may precede a successful completed turn (71.344792ms)
✔ only allowlisted structured diagnostic fields survive without deriving a retry policy (27.926ms)
✔ a failed turn stays failed even if termination flushes completion usage and exits zero (28.137291ms)
✔ valid output and exit zero cannot substitute for an actual completed event (27.594625ms)
✔ a message-only error followed by a hanging process ends at the original deadline (259.258208ms)
✔ a nonzero exit and a completed response with invalid schema both remain failures (84.23775ms)
✔ gateway counts one successful dispatch after diagnostics and retains unknown usage on incomplete failure (111.127541ms)
✔ run failure classification uses terminal AgentFailure and never erases a returned failure (0.166292ms)
✔ Codex process decisions complete the Agent write/verify/evidence loop with isolated transient files (112.922125ms)
✔ Codex only inherits allowlisted environment and uses bounded deduplicated inline images (28.077917ms)
✔ Codex generic notifications do not abort successful terminal events or leak provider text (27.555542ms)
✔ Codex incomplete cannot dispatch or imply Agent success (26.194959ms)
✔ Codex nonzero cannot dispatch or imply Agent success (26.797375ms)
✔ Codex failed-terminal cannot dispatch or imply Agent success (24.792792ms)
✔ Codex bad-event cannot dispatch or imply Agent success (25.545875ms)
✔ Codex external cannot dispatch or imply Agent success (26.803041ms)
✔ Codex stderr-limit cannot dispatch or imply Agent success (26.492042ms)
✔ Codex large cannot dispatch or imply Agent success (25.136333ms)
✔ Codex bad-json cannot dispatch or imply Agent success (25.856917ms)
✔ Codex symlink cannot dispatch or imply Agent success (27.123417ms)
✔ Codex validates original tool arguments and the final schema before dispatch (131.964625ms)
✔ Codex timeout drains final reported usage, deletes files and keeps fixed failure diagnostics (513.857ms)
✔ Codex one-request cancellation is isolated and close drains every outstanding process (53.268792ms)
✔ Codex missing usage counters remain absent and usage hook errors still produce process diagnostics (30.705958ms)
✔ Codex file cleanup failure remains visible to Agent and close without dropping measured usage (33.709625ms)
✔ Codex invalid configuration fails before child launch (0.16075ms)
✔ successful Codex leader cannot leave an owned stdio-independent descendant running (69.18575ms)
✔ cancelled Codex process retains usage emitted while handling SIGTERM (540.421958ms)
✔ gateway close waits for cancellation usage before exposing final metrics (86.671417ms)
✔ named tool_choice rejects a different schema-valid tool (42.259458ms)
✔ tool_choice none permits a text response with tools present (32.749583ms)
✔ tool_choice auto with no selected tools is a stopped text completion (32.312292ms)
✔ timeout terminates a SIGTERM-resistant descendant after its parent exits (1713.352791ms)
✔ Codex transport adapts function decisions, disables external tools, and preserves actual CLI usage (45.415791ms)
✔ Browser Use structured schema and inline images share the same Codex inference transport (31.008833ms)
✔ Codex transport rejects schema-invalid function arguments before the browser runtime sees them (30.565416ms)
✔ automatic tool choice can return ordinary text without inventing an executed call (27.911ms)
✔ unexpected Codex tool activity invalidates the attempt and stops its own process (28.704042ms)
✔ Codex inference deadline and cancellation stop owned child processes without treating them as model answers (207.863375ms)
✔ shared model gateway bridges both framework formats through the same fake Codex process settings (86.689917ms)
✔ Browser Use retains done, true counts, portable history and judge phase when post-task judge times out (236.909916ms)
✔ explicit judge false is preserved and ordinary full return stays distinct from cleanup (65.19875ms)
✔ judge usage contains only reported public response values, and judge exceptions retain their original phase (165.949208ms)
✔ pre-done timeout, post-done exception and non-judge finalization timeout preserve distinct observed phases (495.535542ms)
✔ owned cleanup has a bounded outcome and cannot silently count as a completed run (90.32ms)
✔ Browser Use retains observed download paths when public kill resets browser metadata (55.92675ms)
✔ runner reports independent business success, observed agent done, and full return as separate axes after adapter failure (532.209166ms)
✔ summary retains legacy completedAndPassed while adding explicit axes (0.078417ms)
✔ Tablaze flags reach the official SDK, and proposed finish stays distinct from accepted completion (165.69725ms)
✔ comparison runner stops the owned adapter process group after output overflow (623.665ms)
✔ Browser Use source pin accepts only exact git commits or hashed official archives (55.672417ms)
✔ fixture judges actual records, rejects claimed success, and resets isolated attempts (18.04175ms)
✔ download judge requires the actual matching file, not a download request or a claimed path (4.91325ms)
✔ model gateway enforces equal wire settings and measures returned usage without a paid provider (4.308542ms)
✔ missing model/dependencies stay not_run and never become a competitor failure or a win (29.822958ms)
✔ scripted Tablaze smoke completes diverse real-browser tasks with independent judges (9983.613041ms)
✔ typed backend write, actual browser receipt, and Agent verification complete one reservation (679.166ms)
✔ real navigation while planning removes the restricted tool before its stale decision dispatches (606.977958ms)
✔ handler rechecks a real browser binding after waiting and rejects a stale external write (514.414875ms)
✔ same-origin tenant change cannot resume an uncertain write; matching identity reconciles without replay (1667.552375ms)
✔ CLI rejects a bound checkpoint before launching a browser or restoring its URLs (208.292459ms)
✔ input validation rejects extra fields and attempted context injection before entering a handler (3.864375ms)
✔ global tools work without a browser and receive trusted context, stable callId and a signal (1.142958ms)
✔ caller-selected output may deliberately include a context value (0.706792ms)
✔ published output schema validates the actual MCP envelope and relocated reusable refs (16.597208ms)
✔ write handler exceptions and invalid outputs remain unknown without exposing raw errors (2.578416ms)
✔ trusted read effects remain reads when handlers report failures (0.32625ms)
✔ duplicate registrations, forged handles and reserved names are rejected (0.449875ms)
✔ origins reject wildcard and non-origin restrictions and require a browser binding (0.279583ms)
✔ site filtering includes the exact port and metadata still covers invisible tools (6.468833ms)
✔ closed catalogs release a captured guard once and cannot dispatch handlers (0.293833ms)
✔ catalog context keys stay stable across leases and change when the captured document changes (0.442917ms)
✔ binding changes after catalog preparation refuse dispatch before entering a handler (0.7185ms)
✔ context identity is rechecked after asynchronous validation before the handler starts (0.397792ms)
✔ handler assertCurrent rejects a binding change across await and a started write remains unknown (0.486042ms)
✔ handler assertCurrent also rejects its cancelled dispatch after a noncooperating await (0.375792ms)
✔ a retained execution context is revoked when a successful dispatch has returned (0.339541ms)
✔ independent dispatch cancellation bounds a hanging getContext check before_handler (0.485667ms)
✔ independent dispatch cancellation bounds a hanging guard check before_handler (0.299ms)
✔ independent dispatch cancellation bounds a hanging getContext check after_handler (0.42175ms)
✔ independent dispatch cancellation bounds a hanging guard check after_handler (0.417583ms)
✔ stable contract hashes exclude opaque context values while including schemas and all registry tools (0.337208ms)
✔ an already established registry rejects base-contract and tenant identity changes (0.237458ms)
✔ context resolver failures expose only a fixed registry error (0.240584ms)
✔ runAgent uses catalog dispatch, preserves callId, and records unknown custom writes for reconciliation (2.57875ms)
✔ resume rejects schema, version, effect, origin, base-contract and same-origin tenant changes before planning (4.045583ms)
✔ reconciliation of an unknown write cannot rebind its checkpoint to another tenant on the same origin (4.376625ms)
✔ runAgent discards a decision when the captured browser binding changes during planning (0.54675ms)
✔ viewport observation reaches the 501st button after scrolling a truncated page (712.456375ms)
✔ scoped observation exposes the selected subtree and resets incompatible diff baselines (204.815625ms)
✔ viewport text clips individual offscreen lines within a single text node (178.901083ms)
✔ viewport observation excludes controls and text clipped by scroll-container ancestry (156.2985ms)
✔ snapshots, extraction, verification and guarded actions share open shadow and slot content (250.461542ms)
✔ hidden hosts, hidden slots and raw form values do not leak through composed readers (278.910791ms)
✔ text matching searches beyond bounded evidence and across composed text-node boundaries (185.93025ms)
✔ explicit hidden accessibility labels still name guarded controls without leaking form values (223.98ms)
✔ link and table extraction report item and character truncation accurately (180.899958ms)
✔ scan-budget exhaustion is explicit in both snapshots and extraction (253.3895ms)
✔ ordinary draggable and aria-labeled divs deliver actual native drop payload (719.921958ms)
✔ mouse-driven custom drag supports the observed source covering its drop target (914.114875ms)
✔ nested scaled iframe drag maps main-page coordinates and preserves native payload (1443.685167ms)
✔ open Shadow DOM drag target hit testing reaches the actual source and destination (722.521166ms)
✔ an overlay appearing after the initial mouse move cannot intercept mousedown (1325.155ms)
✔ an overlay appearing during movement cancels the drag before drop and releases the mouse (921.907666ms)
✔ mouseup RPC rejection is a failed action even if the release was applied; cleanup releases held input (1445.750375ms)
✔ an unreleasable mouse closes only the owned session and reports input cleanup failure (876.172625ms)
✔ schema extraction binds every populated leaf to exact source text and authoritative URLs (15.509791ms)
✔ multiple sources can produce nested data and a checked aggregate without a provider dependency (6.732917ms)
✔ invented sources, URLs, approximate quotes and missing or non-leaf citations are rejected (22.174333ms)
✔ JSON pointers escape literal keys and require evidence for null and empty containers (2.256958ms)
✔ schema types, required fields, formats and local references are enforced without coercion (11.080291ms)
✔ literal schema const data is not interpreted as schema references (4.224459ms)
✔ source and candidate limits fail before accepting output (14.98025ms)
✔ quote presence is distinct from truth and the application can reject unsupported claims (17.137042ms)
✔ cancellation and deadline stop waiting on non-cooperating extractors without retries (16.108375ms)
✔ deterministic fields create typed values and preserve their exact DOM provenance (1.9265ms)
✔ field-map cardinality, optional omissions, strict conversions and observation binding are enforced (20.412541ms)
✔ browser inspector uses composed text, current form state and real page provenance (607.693709ms)
✔ closing a CDP session cancels its unfinished download without closing the external browser (750.209084ms)
✔ a tab created after disposal begins is closed instead of escaping session ownership (720.291042ms)
✔ coordinate clicks never move to a sibling tab when the observed tab closes (755.79375ms)
✔ allowed native form submission preserves one POST with the actual encoded business values (759.201042ms)
✔ allowed 307 redirect preserves the original POST method and body without duplicate dispatch (721.567167ms)
✔ same-origin 302 processes Set-Cookie and a matching workspace restores that cookie and policy (1090.644542ms)
✔ explicit open/goto/new-tab deny the other port before a request or a new tab exists (730.898042ms)
✔ multi-hop HTTP 302 is intercepted before the denied hop during open, goto and new-tab (833.225333ms)
✔ iframe redirect is blocked before its denied server receives a request and counts remain per context (1000.123292ms)
✔ one context blocking a document does not interrupt a different context pending business action (1250.79875ms)
✔ a new popup first-document redirect is blocked even before an ordinary popup snapshot exists (837.731833ms)
✔ page script and meta refresh cannot issue an HTTP document request to a denied origin (1008.89225ms)
✔ an allowed localhost child of the 127.0.0.1 page retains interception through its cross-site redirect (966.775167ms)
✔ non-document fetch and image requests reach the other origin: navigation policy is not a network firewall (683.13225ms)
✔ workspace policy changes, missing hashes, policy removal and newly added policy reject before browser launch (0.7225ms)
✔ CDP with configured navigation policy is rejected by the constructor without a connection attempt (51.603917ms)
✔ absent navigation policy retains ordinary browsing and does not attach a policy hash or diagnostics (580.419208ms)
✔ doctor validates and summarizes canonical policy without launching or printing origins (813.155125ms)
✔ policy file parsing rejects unsafe shapes, origin syntax, unknown fields and raw error disclosure (2990.449959ms)
✔ policy reader enforces a 64 KiB file bound and refuses directories and missing files (792.317959ms)
✔ policy reader refuses a FIFO without waiting for a writer (205.362416ms)
✔ policy and external CDP or setup are rejected before reading the policy path (797.020834ms)
✔ malformed private checkpoints do not disclose JSON excerpts before policy restoration (640.224709ms)
✔ MCP startup accepts a valid policy and exposes tools without launching a browser (212.204958ms)
✔ empty CLI checkpoints retain policy identity and reject changed, removed or legacy policy on resume (1276.532083ms)
✔ dispose closes its already-owned browser while policy bootstrap is still awaiting a CDP response (2.365875ms)
✔ an in-flight real navigation reports a safe policy fault instead of the transport URL/error (622.813666ms)
✔ policy failure after a real write retains partial action evidence and skips the next mutation (617.880458ms)
✔ undefined disables policy while configured empty policy permits only HTTP(S) (0.817041ms)
✔ explicit empty allowlist denies all targets and differs from an absent allowlist (0.339834ms)
✔ exact origin matching preserves scheme, port and host boundaries with blocked precedence (0.250083ms)
✔ WHATWG normalization covers case, default ports, Unicode hostnames, IPv4 and IPv6 (0.113875ms)
✔ canonical identity is versioned, stable across equivalent ordering and does not collapse deny-all (0.22575ms)
✔ compiled policy and copied arrays cannot be mutated and detached methods retain their policy (0.228625ms)
✔ only origins are accepted as rules; URL parser repairs and wildcard guesses are rejected (1.039917ms)
✔ strict bounded configuration rejects unknown fields, wrong types, sparse arrays and accessors (0.697333ms)
✔ combined allow/block input is limited to 200 entries before normalization or deduplication (0.611666ms)
✔ invalid configuration, malformed navigation and denied URLs never surface private input (0.378542ms)
✔ pre-cancelled open does not start a browser or create a session (3.510625ms)
✔ cancellation during shared browser startup returns promptly and leaves it available (586.125875ms)
✔ a context that arrives after cancellation is closed without touching a sibling session (590.473ms)
✔ page creation cancellation closes the owned context and any late page (636.02125ms)
✔ navigation cancellation removes only the opening attempt and drains its resources (690.948417ms)
✔ snapshot cancellation cannot register a completed late observation as a live session (726.459542ms)
✔ MCP tab_open propagates cancellation to the engine while sibling sessions survive (651.655834ms)
✔ CDP cancellation cleans a late owned page and leaves independent pages and browser alive (1025.245791ms)
✔ dispose is bounded while an isolated context acquisition has not returned (2470.77125ms)
✔ dispose is bounded while an isolated page acquisition has not returned (2608.996875ms)
✔ dispose reports context close failures rather than claiming complete cleanup (612.676542ms)
✔ dispose retains an existing session cleanup after close removes the session from its map (2450.795917ms)
✔ CDP disposal reports an unidentified pending page and closes it after late arrival (2703.330708ms)
✔ planner removes only exact plain duplicates while preserving errors and distinct or annotated text (10.696458ms)
✔ planner retains text-only fallback and near-duplicate serialization (0.261625ms)
✔ projection preserves image order and complete assistant/tool groups with and without image support (0.514ms)
✔ projection leaves successful evidence, full history, events, and saved checkpoints intact (3.767542ms)
✔ default stay keeps the opener active and completes remaining actions (726.317083ms)
✔ follow-single returns popup observation and truthfully skips remaining old-ref actions (877.82125ms)
✔ multiple candidates are retained without choosing the first or stealing focus (864.527792ms)
✔ a late popup is owned but never retroactively changes the active tab (1501.062ms)
✔ a different owned background opener cannot enter the foreground action window (942.893125ms)
✔ iframe-triggered unique popup is associated with the observed owning page (909.0015ms)
✔ about:blank popups return bounded observations without waiting for eventual navigation (881.877666ms)
✔ popup collection consumes the existing batch deadline and never retries the input (681.79675ms)
✔ workspace preserves policy and old version-one records restore the stay default (2103.664334ms)
✔ failed workspace restoration rolls back policy as well as newly owned sessions (575.432125ms)
✔ stdio handshake exposes the bounded browser tools and structured session errors (216.313375ms)
✔ stdio tools drive isolated Chromium and return browser evidence (889.327458ms)
✔ SDK cancellation stops the browser batch before its later side effect (917.502625ms)
✔ CDP connection errors never return endpoint credentials (231.226292ms)
✔ a total batch budget ends a longer individual wait and reports skipped work (924.541458ms)
✔ CLI help, version and doctor are usable without launching a browser (861.652291ms)
✔ provider selection preserves required explicit model and rejects mismatched options before execution (3063.226708ms)
✔ new provider-only options cannot start the default MCP server (795.017417ms)
✔ Anthropic CLI forwards its explicit model, native tool schema, key and output cap; retains real usage (222.526792ms)
✔ Ollama CLI uses native requests, optional num_predict and no invented token counters (441.400416ms)
✔ explicit compatible provider keeps the previous endpoint and key contract (241.001625ms)
✔ Codex CLI uses the chosen executable and explicit model/effort without a model endpoint (665.971458ms)
✔ Codex executable defaults to codex on PATH without inventing a reasoning setting (675.100209ms)
✔ Codex failures expose fixed diagnostics without stderr, credentials or provider error text (665.951916ms)
✔ Codex provider integrates start-url, actual browser verification and normal cleanup (1343.619292ms)
✔ Codex checkpoint resume preserves prior budgets and still requires provider/model on each invocation (900.092833ms)
✔ bound checkpoints are rejected before any Codex process or browser restoration (232.216709ms)
✔ SIGTERM cancels an in-flight Codex planner and terminates its child (637.094458ms)
✔ anthropic: native wire request completes a verified real MCP tool loop (41.463583ms)
✔ anthropic: images and errors preserve paired repeated-name tool results without mutating history (5.264291ms)
✔ anthropic: text-only selection omits image bytes and native defaults remain explicit (2.083125ms)
✔ anthropic: malformed native decisions, mixed controls and unknown control fields fail safely (8.5795ms)
✔ anthropic: HTTP diagnostics discard private error bodies and do not retry (1.962917ms)
✔ anthropic: native response size is bounded before dropping unknown fields (1.794666ms)
✔ anthropic: cancellation interrupts a pending native response body (3.377708ms)
✔ anthropic: actual missing or invalid usage counters stay absent (1.368458ms)
✔ anthropic: malformed JSON and interrupted bodies retain distinct safe response failures (52.677708ms)
✔ anthropic: unpaired history is rejected before any native request (0.539167ms)
✔ ollama: native wire request completes a verified real MCP tool loop (4.949834ms)
✔ ollama: images and errors preserve paired repeated-name tool results without mutating history (1.075792ms)
✔ ollama: text-only selection omits image bytes and native defaults remain explicit (0.749875ms)
✔ ollama: malformed native decisions, mixed controls and unknown control fields fail safely (3.460833ms)
✔ ollama: HTTP diagnostics discard private error bodies and do not retry (1.314375ms)
✔ ollama: native response size is bounded before dropping unknown fields (0.974875ms)
✔ ollama: cancellation interrupts a pending native response body (0.985958ms)
✔ ollama: actual missing or invalid usage counters stay absent (0.693666ms)
✔ ollama: malformed JSON and interrupted bodies retain distinct safe response failures (53.037833ms)
✔ ollama: unpaired history is rejected before any native request (0.262208ms)
✔ native adapters validate configuration and use only their configured/default native endpoint (0.580959ms)
✔ native factory header validation never includes private names, values or keys in exceptions (0.568167ms)
✔ spawned native-provider CLI rejects an invalid environment key without echoing it to stderr (411.994916ms)
✔ real MCP form fill/save verifies current refs and visible status without another snapshot (774.838417ms)
✔ read-only ref checks preserve actionability and can read a consumed current revision (255.247916ms)
✔ ref checks support textarea, select, empty values and open shadow controls (275.528167ms)
✔ MCP and engine reject missing/ambiguous targets and a ref without snapshot_id (176.075458ms)
✔ new observations and other sessions reject old snapshot identities; unknown refs fail (336.637541ms)
✔ replaced, renamed, hidden and navigated targets are rejected without selector fallback (928.465ms)
✔ parallel status waits cannot preserve ref evidence from a replaced node or old document (819.997459ms)
✔ a later status cannot keep a previously matching value after the same node changes (406.029541ms)
✔ iframe refs retain their frame identity and reject detached frame evidence (340.22275ms)
✔ sensitive and non-form targets fail without leaking values; evidence remains bounded (345.061209ms)
✔ queued verification waits for the action batch and shares the original readonly revision (424.020209ms)
✔ MCP schema extraction returns typed shadow/form data and keeps observed references usable (580.35825ms)
✔ MCP preserves structured validation error codes and pointer issues without raw page values (193.245667ms)
✔ MCP rejects sensitive, hidden, raw-form and truncated fields without partial extracted data (208.630166ms)
✔ MCP extraction follows the explicitly observed frame and cites its own URL (215.217667ms)
✔ MCP bounds field counts and rejects unsupported field options before extraction (166.921542ms)
✔ popup tabs stay owned, switching invalidates references and last close ends session (574.695708ms)
✔ navigation and explicit tabs preserve session state and handle browser history (363.946416ms)
✔ upload reads an explicit local fixture and download produces the actual file (395.491708ms)
✔ native confirm and prompt use one-shot policies and disclose actual results (457.324292ms)
✔ storage state restores cookies and localStorage across independent engine lifetimes (1077.696ms)
✔ hover and double-click execute guarded references, coordinates reject offscreen input (310.859083ms)
✔ CSS-pixel screenshots support actual canvas coordinate input (216.622209ms)
ℹ tests 353
ℹ suites 0
ℹ pass 353
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 152475.088417
