
> 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 (1356.917541ms)
✔ PDF metadata stays bound to its captured tab through asynchronous artifact persistence (703.192834ms)
✔ drag revalidates observed targets after asynchronous geometry reads before mouse input (686.167792ms)
✔ upload chooser revalidates its visible trigger after asynchronous local file checks (553.966792ms)
✔ newly discovered scroll containers retain node identity guards when replaced (608.733667ms)
✔ scrollable containers expose ranges and accept guarded horizontal and vertical input (547.432625ms)
✔ dragging emits the browser drop event and delivers the real data payload (718.283541ms)
✔ hidden file chooser transfers explicit local bytes without a hidden DOM reference (743.048541ms)
✔ PDF exports real private artifacts with verifiable digest and source (629.612542ms)
✔ workspace roundtrip restores isolated durable storage, owned tabs, and active selection (1977.694125ms)
✔ workspace validation precedes navigation and failed restore rolls back owned sessions (948.636166ms)
✔ per-decision catalogs hide execution metadata and context changes skip queued writes (13.385875ms)
✔ a context change during planning or write-ahead persistence never executes the stale handler (2.66525ms)
✔ trusted write effects retain ambiguity for thrown, malformed, error, and unknown outcomes (8.566833ms)
✔ finish rechecks its lease and cannot reuse verification from the old context (2.02125ms)
✔ a context change between closed and newly prepared leases invalidates old verification (4.00875ms)
✔ only a confirmed session close preserves current evidence across context changes and later writes still invalidate it (4.376708ms)
✔ bound checkpoints reject missing runtimes and changed registry or same-origin tenant identities (1.123167ms)
✔ version-one and version-two checkpoints explicitly migrate unbound and cannot gain a registry (1.273375ms)
✔ resume rechecks a pending call effect rather than trusting a tampered mutating flag (1.147458ms)
✔ initialization uses the bound dispatcher and each lease closes before terminal persistence (2.231875ms)
✔ a prepared lease that arrives after cancellation is still closed without planning (3.066042ms)
✔ catalog release failure or cancellation cannot turn accepted verification into success (2.288791ms)
✔ run CLI completes a scripted model task through real Chrome and prints a bounded report (936.737209ms)
✔ run CLI returns human input with exit 2 and an explicit failure with exit 1 (424.518792ms)
✔ run CLI does not accept an unverified model completion and enforces the step limit (232.932708ms)
✔ CLI rejects agent-only options outside run and validates required arguments and limits (2339.486417ms)
✔ CLI reports cleanup failure separately from an already verified successful agent result (756.227209ms)
✔ restoration abort handles a rejected asynchronous dispose without an unhandled rejection (324.723959ms)
✔ CLI publishes bounded planner diagnostics without exposing endpoint, key, or upstream error text (217.831125ms)
✔ HTTP planner failures expose only status and fixed classification without enabling default retries (20.05425ms)
✔ unknown exceptions cannot inject diagnostic codes and custom retry policy remains authoritative (2.718166ms)
✔ transport and malformed response diagnostics remain distinct and omit raw response data (3.41875ms)
✔ catalog failure is distinguishable before any planner or mutation runs (1.642708ms)
✔ application callbacks are classified without recursively notifying a broken event hook (5.325125ms)
✔ adapter usage callback failures retain their application phase and original retry policy input (1.342125ms)
✔ persistence failures prevent mutation or successful completion without changing resume schema (2.015292ms)
✔ a noncooperating checkpoint callback receives a bounded persistence-timeout diagnosis (5002.632542ms)
✔ ordinary MCP explicit unknown write persists uncertainty, skips the batch and cannot replay on resume (17.460542ms)
▶ explicit uncertainty overrides a superficially successful write in legacy and bound MCP execution
  ✔ ordinary MCP write annotation (1.624417ms)
  ✔ bound trusted write metadata (2.851666ms)
✔ explicit uncertainty overrides a superficially successful write in legacy and bound MCP execution (4.668667ms)
▶ ordinary pre-write failures and non-protocol markers remain recoverable by fresh observation
  ✔ absent (1.68025ms)
  ✔ false (2.183167ms)
  ✔ string true (1.548417ms)
  ✔ nested payload (1.450042ms)
  ✔ text-only JSON (1.637209ms)
✔ ordinary pre-write failures and non-protocol markers remain recoverable by fresh observation (9.621208ms)
▶ read-only metadata and trusted not-started outcomes do not invent mutation ambiguity
  ✔ ordinary read (1.080875ms)
  ✔ bound read despite misleading write annotation (0.894209ms)
  ✔ not_started overrides a contradictory result marker (0.659583ms)
✔ read-only metadata and trusted not-started outcomes do not invent mutation ambiguity (2.800459ms)
▶ failure after an explicit unknown result cannot erase the persisted reconciliation requirement
  ✔ after_tool persistence (1.322ms)
  ✔ tool_result callback (1.147791ms)
✔ failure after an explicit unknown result cannot erase the persisted reconciliation requirement (2.556083ms)
✔ explicit initialization and popup following complete a real workflow without executing remaining origin calls (1247.946958ms)
✔ cancellation during terminal persistence cannot return succeeded with accepted evidence (5.151792ms)
✔ deadline expiry during terminal persistence is reported and returned elapsed time includes the wait (153.613708ms)
✔ returned checkpoint accounts for final persistence time before a later resume (43.678708ms)
✔ resume retains application restrictions unless the caller explicitly replaces them (1.210084ms)
✔ resume cannot silently remove a previously required application completion policy (1.549333ms)
✔ decision checkpoint resume preserves counters and IDs but requires fresh verification (17.862667ms)
✔ an interrupted maybe-applied mutation blocks every resumed planner until trusted reconciliation (2.845375ms)
✔ failed checkpoint persistence prevents dispatch and never retries a tool (3.37475ms)
✔ restored sessions reject old session IDs and refs until a fresh snapshot is obtained (2.558834ms)
✔ untrusted checkpoint shape, protocol pairing, IDs, and budgets are validated before execution (2.489916ms)
✔ stall detection uses repeated actual outcomes, provides one replan, then stops (2.051375ms)
✔ bounded retry and fallback only repeat planning; tools with uncertain effects are not retried (3.9295ms)
✔ opt-in compaction preserves complete tool groups and real verification evidence (8.401375ms)
✔ HTTP adapter reports only actual provider usage and exposes transient status classification (10.464ms)
✔ separate CLI processes restore private storage and fresh sessions without replaying completed writes (1991.994833ms)
✔ interrupted external side effect requires explicit reconciliation before CLI restores or plans (1774.057625ms)
✔ an exhausted saved CLI budget performs no page restoration or model request (1227.247708ms)
✔ explicit initialization traverses actual MCP before planning, with step-zero write-ahead and real evidence (13.2835ms)
✔ start URL is explicit and invalid targets fail before catalog, tools, or planning (0.697209ms)
✔ initialization obeys tool, history, deadline, and cancellation budgets without planning (32.821042ms)
✔ unknown initial effects require reconciliation and the initializer is never replayed after it (1.015833ms)
✔ an initializer that was never registered can start once when its saved prerequisite is supplied (0.41325ms)
✔ settled initialization survives resume and compaction with unique subsequent IDs (1.848167ms)
✔ version-one checkpoints migrate explicitly and initialization manifests reject inconsistent history (0.822417ms)
✔ write-ahead persistence failure never dispatches or replays the registered initializer (0.391208ms)
✔ replan_required skips later tool calls and still requires separate verification (0.33325ms)
✔ CLI start URL reaches real Chrome before its first model request and resumes without another initializer (1966.261ms)
✔ CLI accepts a browser-wide popup policy but restricts start URL to run and validates both (1181.69075ms)
✔ agent drives real MCP browser tools and requires verified form state before finishing (736.604542ms)
✔ MCP schema rejects malformed arguments before side effects, then planner can recover (2.616292ms)
✔ stale-reference result stops the batch and is returned for replanning without retry (1.55975ms)
✔ failed checks and isError=true cannot satisfy completion even when passed=true (2.293833ms)
✔ completion rejects stale evidence, another session, invented IDs, and contradicted checks (5.438792ms)
✔ application task-specific acceptance can reject a technically passing unrelated check (0.997166ms)
✔ tool and planning budgets are bounded; invalid decisions never dispatch a tool (1.559084ms)
✔ deadline bounds a non-cooperating planner and abort propagates into an in-flight tool (22.387708ms)
✔ MCP cancellation reaches the actual server request signal (1.224542ms)
✔ HTTP planner preserves MCP image evidence and structured errors without real model calls (10.54475ms)
✔ HTTP planner rejects malformed model arguments, mixed finish calls, and oversized output (0.819ms)
✔ an installed completion policy must explicitly return true (2.1755ms)
✔ history budget stops later mutations within the same planning decision (0.509833ms)
✔ cancellation before the dispatch microtask never starts a custom tool (0.23075ms)
✔ binding is frozen trusted metadata; observations and DOM edits do not consume it or hold the queue (638.581791ms)
✔ same-URL reload invalidates a document lease even if the cached navigation counter is delayed (631.049625ms)
✔ history routes and tab switch ABA invalidate bindings permanently (666.372375ms)
✔ a navigation during the asynchronous assertion cannot produce a valid stale result (598.042583ms)
✔ signal, explicit close, session close and engine disposal revoke leases without exposing browser handles (754.530041ms)
✔ cancelled acquisition releases a real late Document handle without affecting the session (614.224667ms)
✔ pre-cancelled acquisition creates no remote handle and opaque origins are refused (556.431292ms)
✔ engine disposal revokes an in-flight acquisition and releases its late handle (611.702917ms)
✔ independent CDP navigation invalidates a lease and lease cleanup leaves the external browser alive (647.7725ms)
✔ hotel form batch navigates and independent checks verify real URL and DOM (884.860458ms)
✔ snapshots obey budgets and do not include password or hidden values (392.060167ms)
✔ a newer snapshot rejects the older revision before input (190.562666ms)
✔ a replaced DOM node with an identical label is rejected (195.359167ms)
✔ changed target semantics are rejected before a click (194.639625ms)
✔ overlay blocks input and retry requires fresh observation (1264.422375ms)
✔ semantic change during actionability waiting prevents the delayed click (465.609459ms)
✔ a base URL change invalidates the observed resolved link destination (216.502083ms)
✔ value verification never leaks a password substituted after a field-type read (299.810125ms)
✔ batch preserves completed work and reports failed and skipped steps (223.208667ms)
✔ incremental snapshots expose added and removed controls (321.49425ms)
✔ open shadow controls and explicitly selected frame controls execute (353.934041ms)
✔ sessions isolate storage and reject cross-session references (442.680625ms)
✔ concurrent operations serialize within one session (241.809333ms)
✔ capture contains a real JPEG and closing removes a session (208.939875ms)
✔ CDP disposal closes only owned pages and preserves the independent browser and existing state (928.718291ms)
✔ stream and item error diagnostics may precede a successful completed turn (71.325916ms)
✔ only allowlisted structured diagnostic fields survive without deriving a retry policy (28.415875ms)
✔ a failed turn stays failed even if termination flushes completion usage and exits zero (27.508959ms)
✔ valid output and exit zero cannot substitute for an actual completed event (26.390292ms)
✔ a message-only error followed by a hanging process ends at the original deadline (256.004209ms)
✔ a nonzero exit and a completed response with invalid schema both remain failures (64.215083ms)
✔ gateway counts one successful dispatch after diagnostics and retains unknown usage on incomplete failure (112.690833ms)
✔ run failure classification uses terminal AgentFailure and never erases a returned failure (0.159125ms)
✔ Codex process decisions complete the Agent write/verify/evidence loop with isolated transient files (110.180875ms)
✔ Codex only inherits allowlisted environment and uses bounded deduplicated inline images (27.866667ms)
✔ Codex generic notifications do not abort successful terminal events or leak provider text (28.891791ms)
✔ Codex incomplete cannot dispatch or imply Agent success (26.576333ms)
✔ Codex nonzero cannot dispatch or imply Agent success (27.173625ms)
✔ Codex failed-terminal cannot dispatch or imply Agent success (26.129041ms)
✔ Codex bad-event cannot dispatch or imply Agent success (25.298041ms)
✔ Codex external cannot dispatch or imply Agent success (26.234625ms)
✔ Codex stderr-limit cannot dispatch or imply Agent success (27.280959ms)
✔ Codex large cannot dispatch or imply Agent success (26.660167ms)
✔ Codex bad-json cannot dispatch or imply Agent success (25.355459ms)
✔ Codex symlink cannot dispatch or imply Agent success (28.447042ms)
✔ Codex validates original tool arguments and the final schema before dispatch (137.11025ms)
✔ Codex timeout drains final reported usage, deletes files and keeps fixed failure diagnostics (509.219833ms)
✔ Codex one-request cancellation is isolated and close drains every outstanding process (56.388917ms)
✔ Codex missing usage counters remain absent and usage hook errors still produce process diagnostics (31.616083ms)
✔ Codex file cleanup failure remains visible to Agent and close without dropping measured usage (32.270375ms)
✔ Codex invalid configuration fails before child launch (0.134458ms)
✔ successful Codex leader cannot leave an owned stdio-independent descendant running (69.224334ms)
✔ cancelled Codex process retains usage emitted while handling SIGTERM (533.657625ms)
✔ gateway close waits for cancellation usage before exposing final metrics (85.975625ms)
✔ named tool_choice rejects a different schema-valid tool (41.344042ms)
✔ tool_choice none permits a text response with tools present (30.567333ms)
✔ tool_choice auto with no selected tools is a stopped text completion (31.824875ms)
✔ timeout terminates a SIGTERM-resistant descendant after its parent exits (1711.866959ms)
✔ Codex transport adapts function decisions, disables external tools, and preserves actual CLI usage (46.182791ms)
✔ Browser Use structured schema and inline images share the same Codex inference transport (32.416667ms)
✔ Codex transport rejects schema-invalid function arguments before the browser runtime sees them (32.312834ms)
✔ automatic tool choice can return ordinary text without inventing an executed call (29.420667ms)
✔ unexpected Codex tool activity invalidates the attempt and stops its own process (28.746667ms)
✔ Codex inference deadline and cancellation stop owned child processes without treating them as model answers (207.936209ms)
✔ shared model gateway bridges both framework formats through the same fake Codex process settings (86.103292ms)
✔ Browser Use retains done, true counts, portable history and judge phase when post-task judge times out (239.822292ms)
✔ explicit judge false is preserved and ordinary full return stays distinct from cleanup (61.594875ms)
✔ judge usage contains only reported public response values, and judge exceptions retain their original phase (167.389792ms)
✔ pre-done timeout, post-done exception and non-judge finalization timeout preserve distinct observed phases (485.631334ms)
✔ owned cleanup has a bounded outcome and cannot silently count as a completed run (96.239583ms)
✔ Browser Use retains observed download paths when public kill resets browser metadata (60.76625ms)
✔ runner reports independent business success, observed agent done, and full return as separate axes after adapter failure (548.213166ms)
✔ summary retains legacy completedAndPassed while adding explicit axes (0.072625ms)
✔ Tablaze flags reach the official SDK, and proposed finish stays distinct from accepted completion (171.194666ms)
✔ comparison runner stops the owned adapter process group after output overflow (645.347875ms)
✔ Browser Use source pin accepts only exact git commits or hashed official archives (54.543292ms)
✔ fixture judges actual records, rejects claimed success, and resets isolated attempts (17.67425ms)
✔ download judge requires the actual matching file, not a download request or a claimed path (4.638292ms)
✔ model gateway enforces equal wire settings and measures returned usage without a paid provider (3.986584ms)
✔ missing model/dependencies stay not_run and never become a competitor failure or a win (32.883625ms)
✔ scripted Tablaze smoke completes diverse real-browser tasks with independent judges (15163.588625ms)
✔ typed backend write, actual browser receipt, and Agent verification complete one reservation (637.413416ms)
✔ real navigation while planning removes the restricted tool before its stale decision dispatches (638.001208ms)
✔ handler rechecks a real browser binding after waiting and rejects a stale external write (625.47075ms)
✔ same-origin tenant change cannot resume an uncertain write; matching identity reconciles without replay (1676.844791ms)
✔ CLI rejects a bound checkpoint before launching a browser or restoring its URLs (209.302708ms)
✔ input validation rejects extra fields and attempted context injection before entering a handler (3.536042ms)
✔ global tools work without a browser and receive trusted context, stable callId and a signal (1.017542ms)
✔ caller-selected output may deliberately include a context value (0.605625ms)
✔ published output schema validates the actual MCP envelope and relocated reusable refs (16.112791ms)
✔ write handler exceptions and invalid outputs remain unknown without exposing raw errors (2.781666ms)
✔ trusted read effects remain reads when handlers report failures (0.372667ms)
✔ duplicate registrations, forged handles and reserved names are rejected (0.456625ms)
✔ origins reject wildcard and non-origin restrictions and require a browser binding (0.268458ms)
✔ site filtering includes the exact port and metadata still covers invisible tools (6.298542ms)
✔ closed catalogs release a captured guard once and cannot dispatch handlers (0.295792ms)
✔ catalog context keys stay stable across leases and change when the captured document changes (0.525375ms)
✔ binding changes after catalog preparation refuse dispatch before entering a handler (0.770291ms)
✔ context identity is rechecked after asynchronous validation before the handler starts (0.39875ms)
✔ handler assertCurrent rejects a binding change across await and a started write remains unknown (0.42825ms)
✔ handler assertCurrent also rejects its cancelled dispatch after a noncooperating await (0.264458ms)
✔ a retained execution context is revoked when a successful dispatch has returned (0.293541ms)
✔ independent dispatch cancellation bounds a hanging getContext check before_handler (0.43625ms)
✔ independent dispatch cancellation bounds a hanging guard check before_handler (0.269042ms)
✔ independent dispatch cancellation bounds a hanging getContext check after_handler (0.539167ms)
✔ independent dispatch cancellation bounds a hanging guard check after_handler (0.52075ms)
✔ stable contract hashes exclude opaque context values while including schemas and all registry tools (0.361583ms)
✔ an already established registry rejects base-contract and tenant identity changes (0.1875ms)
✔ context resolver failures expose only a fixed registry error (0.226458ms)
✔ runAgent uses catalog dispatch, preserves callId, and records unknown custom writes for reconciliation (2.548375ms)
✔ resume rejects schema, version, effect, origin, base-contract and same-origin tenant changes before planning (6.878875ms)
✔ reconciliation of an unknown write cannot rebind its checkpoint to another tenant on the same origin (1.052959ms)
✔ runAgent discards a decision when the captured browser binding changes during planning (0.532292ms)
✔ viewport observation reaches the 501st button after scrolling a truncated page (729.887209ms)
✔ scoped observation exposes the selected subtree and resets incompatible diff baselines (217.651292ms)
✔ viewport text clips individual offscreen lines within a single text node (181.4815ms)
✔ viewport observation excludes controls and text clipped by scroll-container ancestry (183.711875ms)
✔ snapshots, extraction, verification and guarded actions share open shadow and slot content (262.235209ms)
✔ hidden hosts, hidden slots and raw form values do not leak through composed readers (288.317459ms)
✔ text matching searches beyond bounded evidence and across composed text-node boundaries (179.240375ms)
✔ explicit hidden accessibility labels still name guarded controls without leaking form values (237.50075ms)
✔ link and table extraction report item and character truncation accurately (191.904375ms)
✔ scan-budget exhaustion is explicit in both snapshots and extraction (279.919958ms)
✔ ordinary draggable and aria-labeled divs deliver actual native drop payload (746.230833ms)
✔ mouse-driven custom drag supports the observed source covering its drop target (956.185417ms)
✔ nested scaled iframe drag maps main-page coordinates and preserves native payload (1503.351375ms)
✔ open Shadow DOM drag target hit testing reaches the actual source and destination (720.115333ms)
✔ an overlay appearing after the initial mouse move cannot intercept mousedown (1354.766792ms)
✔ an overlay appearing during movement cancels the drag before drop and releases the mouse (903.9215ms)
✔ mouseup RPC rejection is a failed action even if the release was applied; cleanup releases held input (1501.146625ms)
✔ an unreleasable mouse closes only the owned session and reports input cleanup failure (889.38625ms)
✔ schema extraction binds every populated leaf to exact source text and authoritative URLs (15.394042ms)
✔ multiple sources can produce nested data and a checked aggregate without a provider dependency (6.576333ms)
✔ invented sources, URLs, approximate quotes and missing or non-leaf citations are rejected (22.336125ms)
✔ JSON pointers escape literal keys and require evidence for null and empty containers (2.427166ms)
✔ schema types, required fields, formats and local references are enforced without coercion (11.062459ms)
✔ literal schema const data is not interpreted as schema references (3.275833ms)
✔ source and candidate limits fail before accepting output (15.321125ms)
✔ quote presence is distinct from truth and the application can reject unsupported claims (17.151541ms)
✔ cancellation and deadline stop waiting on non-cooperating extractors without retries (17.478167ms)
✔ deterministic fields create typed values and preserve their exact DOM provenance (1.983875ms)
✔ field-map cardinality, optional omissions, strict conversions and observation binding are enforced (21.392291ms)
✔ browser inspector uses composed text, current form state and real page provenance (635.288ms)
✔ find traverses actual virtualized rows and returns a fresh actionable viewport ref (4706.376958ms)
✔ MCP find supports a selected iframe and preserves current snapshot semantics (831.546959ms)
✔ closing a CDP session cancels its unfinished download without closing the external browser (779.82725ms)
✔ a tab created after disposal begins is closed instead of escaping session ownership (779.355375ms)
✔ coordinate clicks never move to a sibling tab when the observed tab closes (821.732334ms)
✔ allowed native form submission preserves one POST with the actual encoded business values (790.198458ms)
✔ allowed 307 redirect preserves the original POST method and body without duplicate dispatch (800.522875ms)
✔ same-origin 302 processes Set-Cookie and a matching workspace restores that cookie and policy (1136.091833ms)
✔ explicit open/goto/new-tab deny the other port before a request or a new tab exists (748.116458ms)
✔ multi-hop HTTP 302 is intercepted before the denied hop during open, goto and new-tab (978.203792ms)
✔ iframe redirect is blocked before its denied server receives a request and counts remain per context (1000.514208ms)
✔ one context blocking a document does not interrupt a different context pending business action (1249.305125ms)
✔ a new popup first-document redirect is blocked even before an ordinary popup snapshot exists (860.308125ms)
✔ page script and meta refresh cannot issue an HTTP document request to a denied origin (1149.046625ms)
✔ an allowed localhost child of the 127.0.0.1 page retains interception through its cross-site redirect (975.633042ms)
✔ non-document fetch and image requests reach the other origin: navigation policy is not a network firewall (586.183708ms)
✔ workspace policy changes, missing hashes, policy removal and newly added policy reject before browser launch (0.693833ms)
✔ CDP with configured navigation policy is rejected by the constructor without a connection attempt (52.543375ms)
✔ absent navigation policy retains ordinary browsing and does not attach a policy hash or diagnostics (529.914292ms)
✔ doctor validates and summarizes canonical policy without launching or printing origins (789.365542ms)
✔ policy file parsing rejects unsafe shapes, origin syntax, unknown fields and raw error disclosure (2993.379833ms)
✔ policy reader enforces a 64 KiB file bound and refuses directories and missing files (782.012292ms)
✔ policy reader refuses a FIFO without waiting for a writer (200.632042ms)
✔ policy and external CDP or setup are rejected before reading the policy path (787.136041ms)
✔ malformed private checkpoints do not disclose JSON excerpts before policy restoration (598.794666ms)
✔ MCP startup accepts a valid policy and exposes tools without launching a browser (208.408167ms)
✔ empty CLI checkpoints retain policy identity and reject changed, removed or legacy policy on resume (1255.054541ms)
✔ dispose closes its already-owned browser while policy bootstrap is still awaiting a CDP response (2.818459ms)
✔ an in-flight real navigation reports a safe policy fault instead of the transport URL/error (628.372583ms)
✔ policy failure after a real write retains partial action evidence and skips the next mutation (669.066792ms)
✔ undefined disables policy while configured empty policy permits only HTTP(S) (0.674458ms)
✔ explicit empty allowlist denies all targets and differs from an absent allowlist (0.324916ms)
✔ exact origin matching preserves scheme, port and host boundaries with blocked precedence (0.234834ms)
✔ WHATWG normalization covers case, default ports, Unicode hostnames, IPv4 and IPv6 (0.111583ms)
✔ canonical identity is versioned, stable across equivalent ordering and does not collapse deny-all (0.215625ms)
✔ compiled policy and copied arrays cannot be mutated and detached methods retain their policy (0.216917ms)
✔ only origins are accepted as rules; URL parser repairs and wildcard guesses are rejected (1.081917ms)
✔ strict bounded configuration rejects unknown fields, wrong types, sparse arrays and accessors (0.899834ms)
✔ combined allow/block input is limited to 200 entries before normalization or deduplication (0.632ms)
✔ invalid configuration, malformed navigation and denied URLs never surface private input (0.419208ms)
✔ pre-cancelled open does not start a browser or create a session (3.844375ms)
✔ cancellation during shared browser startup returns promptly and leaves it available (632.572ms)
✔ a context that arrives after cancellation is closed without touching a sibling session (596.869125ms)
✔ page creation cancellation closes the owned context and any late page (723.188ms)
✔ navigation cancellation removes only the opening attempt and drains its resources (673.180917ms)
✔ snapshot cancellation cannot register a completed late observation as a live session (748.464625ms)
✔ MCP tab_open propagates cancellation to the engine while sibling sessions survive (757.203083ms)
✔ CDP cancellation cleans a late owned page and leaves independent pages and browser alive (976.7095ms)
✔ dispose is bounded while an isolated context acquisition has not returned (2469.085167ms)
✔ dispose is bounded while an isolated page acquisition has not returned (2512.921333ms)
✔ dispose reports context close failures rather than claiming complete cleanup (615.42925ms)
✔ dispose retains an existing session cleanup after close removes the session from its map (2452.421083ms)
✔ CDP disposal reports an unidentified pending page and closes it after late arrival (2819.23425ms)
✔ planner removes only exact plain duplicates while preserving errors and distinct or annotated text (10.425334ms)
✔ planner retains text-only fallback and near-duplicate serialization (0.382958ms)
✔ projection preserves image order and complete assistant/tool groups with and without image support (0.545583ms)
✔ projection leaves successful evidence, full history, events, and saved checkpoints intact (3.886834ms)
✔ default stay keeps the opener active and completes remaining actions (756.919542ms)
✔ follow-single returns popup observation and truthfully skips remaining old-ref actions (831.3885ms)
✔ multiple candidates are retained without choosing the first or stealing focus (867.8775ms)
✔ a late popup is owned but never retroactively changes the active tab (1550.682458ms)
✔ a different owned background opener cannot enter the foreground action window (928.747666ms)
✔ iframe-triggered unique popup is associated with the observed owning page (874.965416ms)
✔ about:blank popups return bounded observations without waiting for eventual navigation (926.618208ms)
✔ popup collection consumes the existing batch deadline and never retries the input (715.544583ms)
✔ workspace preserves policy and old version-one records restore the stay default (2188.237875ms)
✔ failed workspace restoration rolls back policy as well as newly owned sessions (567.834209ms)
✔ stdio handshake exposes the bounded browser tools and structured session errors (218.050917ms)
✔ stdio tools drive isolated Chromium and return browser evidence (945.786875ms)
✔ SDK cancellation stops the browser batch before its later side effect (860.224708ms)
✔ CDP connection errors never return endpoint credentials (219.479458ms)
✔ a total batch budget ends a longer individual wait and reports skipped work (1023.531708ms)
✔ CLI help, version and doctor are usable without launching a browser (793.450125ms)
✔ provider selection preserves required explicit model and rejects mismatched options before execution (2974.812875ms)
✔ new provider-only options cannot start the default MCP server (776.643166ms)
✔ Anthropic CLI forwards its explicit model, native tool schema, key and output cap; retains real usage (220.4105ms)
✔ Ollama CLI uses native requests, optional num_predict and no invented token counters (428.331625ms)
✔ explicit compatible provider keeps the previous endpoint and key contract (212.260833ms)
✔ Codex CLI uses the chosen executable and explicit model/effort without a model endpoint (646.607792ms)
✔ Codex executable defaults to codex on PATH without inventing a reasoning setting (648.826958ms)
✔ Codex failures expose fixed diagnostics without stderr, credentials or provider error text (641.021709ms)
✔ Codex provider integrates start-url, actual browser verification and normal cleanup (1272.962584ms)
✔ Codex checkpoint resume preserves prior budgets and still requires provider/model on each invocation (894.404166ms)
✔ bound checkpoints are rejected before any Codex process or browser restoration (203.812ms)
✔ SIGTERM cancels an in-flight Codex planner and terminates its child (627.856ms)
✔ anthropic: native wire request completes a verified real MCP tool loop (32.335708ms)
✔ anthropic: images and errors preserve paired repeated-name tool results without mutating history (3.856041ms)
✔ anthropic: text-only selection omits image bytes and native defaults remain explicit (1.24025ms)
✔ anthropic: malformed native decisions, mixed controls and unknown control fields fail safely (6.413791ms)
✔ anthropic: HTTP diagnostics discard private error bodies and do not retry (1.826625ms)
✔ anthropic: native response size is bounded before dropping unknown fields (1.190375ms)
✔ anthropic: cancellation interrupts a pending native response body (1.544833ms)
✔ anthropic: actual missing or invalid usage counters stay absent (1.215583ms)
✔ anthropic: malformed JSON and interrupted bodies retain distinct safe response failures (53.889333ms)
✔ anthropic: unpaired history is rejected before any native request (0.454167ms)
✔ ollama: native wire request completes a verified real MCP tool loop (5.065667ms)
✔ ollama: images and errors preserve paired repeated-name tool results without mutating history (1.158917ms)
✔ ollama: text-only selection omits image bytes and native defaults remain explicit (0.732666ms)
✔ ollama: malformed native decisions, mixed controls and unknown control fields fail safely (2.961459ms)
✔ ollama: HTTP diagnostics discard private error bodies and do not retry (0.8085ms)
✔ ollama: native response size is bounded before dropping unknown fields (0.77975ms)
✔ ollama: cancellation interrupts a pending native response body (0.823875ms)
✔ ollama: actual missing or invalid usage counters stay absent (0.678417ms)
✔ ollama: malformed JSON and interrupted bodies retain distinct safe response failures (54.441292ms)
✔ ollama: unpaired history is rejected before any native request (0.251583ms)
✔ native adapters validate configuration and use only their configured/default native endpoint (0.522ms)
✔ native factory header validation never includes private names, values or keys in exceptions (0.4735ms)
✔ spawned native-provider CLI rejects an invalid environment key without echoing it to stderr (390.313583ms)
✔ real MCP form fill/save verifies current refs and visible status without another snapshot (775.167917ms)
✔ read-only ref checks preserve actionability and can read a consumed current revision (227.406583ms)
✔ ref checks support textarea, select, empty values and open shadow controls (282.810375ms)
✔ MCP and engine reject missing/ambiguous targets and a ref without snapshot_id (187.281208ms)
✔ new observations and other sessions reject old snapshot identities; unknown refs fail (348.293875ms)
✔ replaced, renamed, hidden and navigated targets are rejected without selector fallback (951.978584ms)
✔ parallel status waits cannot preserve ref evidence from a replaced node or old document (753.509ms)
✔ a later status cannot keep a previously matching value after the same node changes (393.63575ms)
✔ iframe refs retain their frame identity and reject detached frame evidence (403.801917ms)
✔ sensitive and non-form targets fail without leaking values; evidence remains bounded (342.147708ms)
✔ queued verification waits for the action batch and shares the original readonly revision (430.218417ms)
✔ scoped credential completes a real login and never appears in normal evidence or artifacts (1043.882833ms)
✔ iframe credentials require both the frame and top-level origin to match (832.329666ms)
✔ MCP uses only the credential alias and reports login evidence without disclosing the value (2748.423292ms)
✔ doctor reports only secret metadata and never launches the configured browser (403.871625ms)
✔ malformed secret configuration fails before doctor or run with one fixed message (1178.053834ms)
✔ CDP and setup reject secrets before reading the configuration path (781.010458ms)
✔ stdio accepts configured aliases and exposes the public action without resolving or launching (209.780917ms)
✔ CLI run saves secret policy identity and requires the same configuration on resume (832.861625ms)
✔ configuration does not read env values until an allowed-origin resolution and supports rotation (5.028208ms)
✔ returned configuration and scopes are frozen, with only explicit artifact opt-in (1.356375ms)
✔ resolvers reject missing, inherited, empty and oversized values without disclosing their source (1.176208ms)
✔ aborted resolution does not read env and an abort during a getter cannot return the value (0.593708ms)
✔ strict JSON shape and shared SDK metadata limits reject malformed private configurations (4.142667ms)
✔ reader enforces the actual 64 KiB boundary, UTF-8 and regular files with fixed errors (0.906458ms)
✔ reader refuses a FIFO without blocking on a writer (3.029375ms)
✔ pre-page bridge exposes only bounded metadata and fills the original control without focusing a decoy (495.915333ms)
✔ old handles fail after identical DOM replacement without filling the replacement (190.511584ms)
✔ opaque HTTP sandbox remains denied despite replacing window.origin in page script (223.4715ms)
✔ captured native primitives withstand later prototype and global replacements (156.208958ms)
✔ bridge property and operation references cannot be replaced by page code (151.54575ms)
✔ invalid bindings, values and unsupported or disabled controls reject before writing (157.637167ms)
✔ an input listener detaching the filled node reports wrote=true and does not dispatch change or fill its replacement (151.317167ms)
✔ same-URL navigation generates a fresh document identity and rejects the prior binding (168.700416ms)
✔ text projection uses lookahead before clipping and never promotes its trailing content (0.910958ms)
✔ raw boundaries remain exact when a secret resembles the replacement marker or overlaps another secret (0.833375ms)
✔ snapshot projection covers fields, options, diffs and metadata without changing reference identity (0.581125ms)
✔ link extraction preserves href-first raw allocation when redaction shrinks the last field (0.168959ms)
✔ link text and table cells redact before their field and aggregate boundaries (0.13925ms)
✔ short secrets cannot expand links or table cells past the aggregate public budget (1.94375ms)
✔ expansion followed by shrinking never reallocates the original reader budget to later fields (0.61225ms)
✔ text extraction restores the original 20000 character boundary (0.090584ms)
✔ selected metadata is copied and recursively redacted with bounded rejection (0.2945ms)
✔ metadata bounds replacement expansion as well as raw input size (45.121292ms)
✔ published and runtime MCP schemas accept only a bounded alias plus observed ref for fill_secret (29.155584ms)
✔ secret failures retain their fixed public code in structured and text MCP results (1.325667ms)
✔ SDK server rejects secret configuration with external CDP before any resolver is called (0.219459ms)
✔ undefined disables secrets; compiled facade is frozen, detached methods work, and values are private (1.871291ms)
✔ strict options reject unknown keys, accessors, prototypes and wrong optional types without invoking getters (0.57225ms)
✔ strict definitions reject invalid names, versions, duplicate aliases, unknown fields and inherited resolvers (0.280583ms)
✔ arrays reject holes, extra properties and getters; alias and configured-origin limits are bounded (0.960834ms)
✔ origin lists require exact nonempty HTTP(S) origins; wildcard, credentials, path and opaque origins are rejected (0.243459ms)
✔ frame and top origins are both required, including port and scheme; scope failure never invokes resolver (0.388083ms)
✔ hash normalizes ordering, duplicates and explicit default top scope; excludes callbacks and resolved values (0.510208ms)
✔ caller mutations cannot change cloned scopes, metadata or captured resolver (0.128208ms)
✔ context is checked before resolving and again before delivering; failed delivery still redacts the resolved value (0.216625ms)
✔ resolver and context exceptions are fixed and never expose raw thrown objects or messages (0.548166ms)
✔ pre-aborted calls invoke neither context nor resolver and never expose the abort reason (0.08675ms)
✔ cancellation bounds a hanging pre-context check and does not enter the resolver later (2.341209ms)
✔ cancellation bounds resolver wait, passes an aborted signal, and registers its late value while store remains alive (0.378ms)
✔ late resolver rejection after cancellation is observed without unhandled rejection (0.222916ms)
✔ cancellation bounds a hanging post-resolution context check while preserving redaction (0.222833ms)
✔ all rotated and cross-alias values remain redacted; duplicate values do not consume retention slots (1.978916ms)
✔ resolved values must be nonempty bounded well-formed strings and errors contain no returned value (0.332875ms)
✔ 128 distinct-value limit fails before delivery without dropping earlier redactions (2.084708ms)
✔ 64 KiB retention limit counts UTF-8 bytes and rejects overflow atomically (9.769125ms)
✔ raw, whitespace-normalized, URI and JSON-escaped variants redact literally and contains follows the same set (0.2635ms)
✔ overlapping secrets use longest simultaneous replacement and never reprocess the replacement marker (0.152583ms)
✔ crossing and chained matches cover every sensitive character in original coordinates (0.181334ms)
✔ public aliases equal to known values are hidden without changing configured resolution or unrelated names (0.157125ms)
✔ a late cancelled value and a newer rotated value are both retained until close (1.671375ms)
✔ padding covers maximum URI and JSON forms; redaction before truncation does not reveal a secret prefix (2.94ms)
✔ close revokes active resolution, aborts the resolver signal and prevents silent redaction bypass afterward (0.349292ms)
✔ MCP schema extraction returns typed shadow/form data and keeps observed references usable (611.8865ms)
✔ MCP preserves structured validation error codes and pointer issues without raw page values (185.313625ms)
✔ MCP rejects sensitive, hidden, raw-form and truncated fields without partial extracted data (227.107667ms)
✔ MCP extraction follows the explicitly observed frame and cites its own URL (226.802667ms)
✔ MCP bounds field counts and rejects unsupported field options before extraction (184.212709ms)
✔ real actions show a non-intercepting pointer on owned pages and rejected refs do not fake a click (3431.86825ms)
✔ operators can disable the visual pointer (653.12625ms)
✔ sessions do not add a pointer by default, even when headed (772.298ms)
✔ popup tabs stay owned, switching invalidates references and last close ends session (631.4875ms)
✔ navigation and explicit tabs preserve session state and handle browser history (341.06275ms)
✔ upload reads an explicit local fixture and download produces the actual file (300.779834ms)
✔ native confirm and prompt use one-shot policies and disclose actual results (430.810917ms)
✔ storage state restores cookies and localStorage across independent engine lifetimes (1157.116083ms)
✔ hover and double-click execute guarded references, coordinates reject offscreen input (321.659084ms)
✔ CSS-pixel screenshots support actual canvas coordinate input (225.055917ms)
ℹ tests 437
ℹ suites 0
ℹ pass 437
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 181431.57725
