
> snake-eyes-editor@0.4.168 test
> node --test tests/*.test.js

✔ audio-only correction is visible beside unchanged video in summary (0.883583ms)
✔ empty tracks report N/A rather than perfect retention (0.084166ms)
✔ dashboard renders exact audio trim result and labels separate columns (61.001667ms)
✔ incremental directory read closes handle when budget is exceeded (1.430584ms)
✔ directory budgets include ignored files and span delivery folders (1.809542ms)
✔ invalid directory limits fail before disk access (0.654958ms)
✔ stable reader counts exact bytes and refuses exhausted aggregate budget (14.216125ms)
✔ stable reader refuses growing, shrinking, replaced and rewritten files (45.271041ms)
✔ stable reader closes descriptor after failed reads and refuses links (10.434791ms)
✔ release audits and inventories refuse excessive directory nesting (106.234625ms)
✔ release inventory cannot hash an old file after its path is replaced (1.191166ms)
✔ distribution audit labels changing text as an issue (2.159958ms)
✔ disconnect cannot clear storage during a queue command (1.367458ms)
✔ folder replacement cannot redirect an active command (0.517583ms)
✔ panel editing and import operations prevent disconnect (1.038959ms)
✔ idle disconnect still clears timer storage and saved token (0.426958ms)
✔ failed activation retains the previous connection and cached history (0.421792ms)
✔ activation locks connection changes until initialization completes (0.444125ms)
✔ successful replacement removes the previous polling timer (0.595666ms)
✔ failed history persistence restores previous active context (0.527209ms)
✔ failed new connection does not replace the saved persistent token (0.498708ms)
✔ failed automatic restore preserves the token for a later retry (0.473917ms)
✔ panel read owns the bridge lock and prevents connection changes until completion (0.418458ms)
✔ panel read refuses active edits and releases only its own lock after failure (0.390958ms)
✔ state writer creates exclusively and verifies publication (1.472792ms)
✔ unreadable inventory cannot trigger an overwrite fallback (1.824292ms)
✔ silent or corrupt writes cannot report success (0.780584ms)
✔ duplicate non-file and incomplete inventories prevent writes (0.987458ms)
✔ path-like or empty filenames fail before filesystem access (3.545125ms)
✔ writer snapshots the value before the first asynchronous filesystem call (0.345625ms)
✔ unlisted created file cannot count as a verified publication (0.303125ms)
✔ existing state is updated without recreating its file (0.276ms)
✔ a file appearing during creation is preserved rather than overwritten (1.283375ms)
✔ wrong created entry identity is rejected before writing (0.418959ms)
✔ exclusive state publication refuses an existing record without changing bytes (0.31975ms)
✔ exclusive publication supports a new record and rejects ambiguous options (0.901209ms)
✔ CRC agrees with standard check vector (0.45175ms)
✔ CCX opens in independent ZIP reader with root manifest, valid CRC and no companion (83.58ms)
✔ identical release bytes produce identical archives (82.244ms)
✔ CCX never overwrites existing output or leaves temporary files (69.21175ms)
✔ changed release bytes fail integrity before publication (11.874167ms)
✔ runtime state rejected even though installation checker permits state (4.116584ms)
✔ manifest requires one Premiere host, matching version and packaged entry point (41.13825ms)
✔ symlink source cannot enter package (6.655416ms)
✔ outputs inside release, symlink parents and incorrect extension are refused (16.901458ms)
✔ ZIP entry guard rejects traversal, duplicates and ambiguous names (0.287541ms)
✔ UTF8 filename and empty file survive independent ZIP reader (114.795541ms)
✔ portable ZIP names reject reserved devices, controls, trailing aliases and invalid Unicode (0.269167ms)
✔ ZIP refuses case, normalization and file-versus-directory collisions (0.304542ms)
✔ manifest changes during audit invalidate the captured release (57.953208ms)
✔ short or corrupted archive writes are refused before publication (124.664541ms)
✔ archive readback errors leave no published output or scratch file (54.034ms)
✔ failed descriptor close still attempts to remove its private temporary file (43.2075ms)
✔ archive verification tolerates valid partial filesystem reads (42.6465ms)
✔ fresh context never reads remembered evidence and preserves current brief (1.65675ms)
✔ stale, future, invalid mode requests rejected (0.147834ms)
✔ request capture binds saved choice and active sequence (0.597583ms)
✔ project switches during capture do not publish (0.446708ms)
✔ neutral transform is identity and neutral LUT uses red-fastest cube ordering (5.700709ms)
✔ all looks produce finite bounded output and preserve endpoints (1.000417ms)
✔ zero saturation produces gray; exposure increases midgray (0.103334ms)
✔ unknown controls, excessive controls, and unverified spaces rejected (0.123667ms)
✔ shot plans retain distinct controls and reject duplicate identities or reversed ranges (0.338208ms)
✔ color build writes LUTs and matching manifest, refuses replacement and validates before writes (50.984375ms)
✔ sparse RGB values never become NaN color output (0.54775ms)
✔ prepared transforms pin validated controls (0.358375ms)
✔ prepared and one-shot transforms agree across controls and RGB values (6.787459ms)
✔ sparse shot lists are rejected instead of producing incomplete plans (0.156667ms)
✔ sparse mask coordinates, polygons, corrections and masks fail validation (0.294292ms)
✔ spatial processing preserves input arrays and finite output (0.67975ms)
✔ CLI refuses explicitly invalid LUT sizes before creating output (33.042ms)
✔ unknown commands extra args and option-shaped arguments fail before publication (199.390916ms)
✔ decimal arguments and identities do not accept coercion or path traversal (0.27ms)
✔ catalog covers actual host commands except dedicated video-export interface (0.842916ms)
✔ required optional and no-argument commands enforce exact arity (0.282875ms)
✔ receipt identifiers and decimal arguments have explicit safe boundaries (0.225833ms)
✔ delivery executes once and uses the stored result on replay (0.59075ms)
✔ missing result after handler entry produces uncertainty and prevents replay (0.14ms)
✔ intent surviving restart before host entry is not blindly executed (0.075208ms)
✔ unavailable result storage retains pending evidence across retries (0.678417ms)
✔ failed intent persistence prevents host entry (0.077542ms)
✔ conflicting command IDs and corrupt results never reach a handler (0.08925ms)
✔ legacy result prevents replay without manufacturing an intent (0.062125ms)
✔ delivery snapshots input and refuses overlapping handler entry (0.095375ms)
✔ delivery status reports uncertain outcome separately from completed execution (0.099ms)
✔ exclusive publication refuses collisions and preserves the existing command (32.503125ms)
✔ killed sender before publication never exposes partial command JSON (48.4425ms)
✔ killed sender after publication never exposes partial command JSON (75.131709ms)
✔ eight independent senders publish unique valid commands (87.637167ms)
✔ invalid JSON input cannot publish a command (28.808208ms)
✔ serialization finishes before final command name becomes visible (16.229708ms)
✔ command serialization is compact, deterministic and preserves complete input (0.952084ms)
✔ queue byte limit measures UTF8 bytes and accepts the exact boundary (0.265417ms)
✔ serialization rejects hooks, sparse data and invalid envelopes before publication (0.138083ms)
✔ deep plans are not expanded with indentation in the queue (58.269708ms)
✔ excessive command nesting fails before queue directories are created (35.297458ms)
✔ unsafe decimal arguments are refused without creating queue state (56.909417ms)
✔ command byte-limit options cannot disable the bound (0.159792ms)
✔ companion self-test succeeds and leaves existing scratch-parent contents alone (216.276667ms)
✔ missing or redirected scratch folders fail without creating or following them (5.721417ms)
✔ worker-start failure is reported and scratch is still cleaned (16.222083ms)
✔ timed-out workers are killed and confirmed closed before cleanup (66.258792ms)
✔ unverified worker termination preserves only its unique scratch directory (99.68875ms)
✔ cleanup failure never produces a healthy report (217.742333ms)
✔ self-test options and CLI usage are strict (29.5175ms)
✔ shipped CLI completes synthetic checks without requiring initialized state (233.310583ms)
✔ exclusive-write failure is visible while independent diagnostics still run (265.477417ms)
✔ timestamps labels warning wording and key order cannot inflate editorial count (1.237583ms)
✔ distinct edits, tracks, timings and sequence identities remain distinct (0.923292ms)
✔ occurrence order and repetitions are preserved in comparison identity (0.103708ms)
✔ equivalent exports cannot reach five-comparison display threshold (0.154667ms)
✔ dashboard deduplicates equivalent exports while preserving purpose (86.499417ms)
✔ invalid excluded test record cannot partially replace last good reports (76.496458ms)
✔ sweep comparison agrees with independent interval oracle across 500 overlapping cases (32.463459ms)
✔ twenty thousand ranges in one 17-hour source retain precise full overlap (42.686208ms)
✔ nested repeated intervals preserve multiplicity instead of double counting (6.899542ms)
✔ read-only status handshake confirms the loaded runtime version (52.16725ms)
✔ stale runtime is reachable but does not pass a version check (20.354959ms)
✔ missing state is not created and redirected directories are refused (1.335958ms)
✔ timeout preserves the harmless diagnostic request instead of racing host consumption (34.347083ms)
✔ partial reply can complete within the bounded wait (65.212584ms)
✔ failed and malformed status replies are not reported as connected (92.590125ms)
✔ invalid wait settings fail before publishing any request (1.3935ms)
✔ wrong identifiers and linked replies cannot satisfy a handshake (181.730292ms)
✔ redirected results directory during a wait is unavailable (29.757292ms)
✔ unsafe version components are refused before publication (6.38825ms)
✔ CLI refuses unexpected arguments without creating state (126.880875ms)
✔ 0.4.63 host implementation preserved except version and tested import receipt handoff (19.1895ms)
✔ current panel boots original controls and additions together without mutation (4.13125ms)
✔ current XML imports keep durable replay protection and exact host arguments (2.939458ms)
✔ current import receipt failure prevents host entry (0.209583ms)
✔ import changes are limited to tested guards and verified receipt publication (0.350916ms)
✔ visible panel version agrees with installed manifests and runtime (0.505917ms)
✔ absent history is not created by inspection (12.45625ms)
✔ queued and uncertain requests are distinguished without exposing payloads (11.878958ms)
✔ recorded results disappear from pending inventory but uncertain results remain (14.628625ms)
✔ conflicting queue payload is flagged even with a successful result (7.724333ms)
✔ malformed and linked evidence cannot be mistaken for no history (62.998291ms)
✔ linked state and child directories are rejected (31.4375ms)
✔ invalid IDs are refused without reaching the filesystem (0.851292ms)
✔ CLI reports corrupt evidence with an error exit and no raw payload (59.298792ms)
✔ falsy and non-object JSON evidence is invalid rather than absent (9.00675ms)
✔ completed request still in the queue remains visible for cleanup diagnosis (1.703083ms)
✔ pending inventory limits aggregate bytes instead of accepting many individually valid files (5.315333ms)
✔ completed result and intent bytes count toward aggregate inventory limit (5.371958ms)
✔ inventory limit validation happens without creating state (1.326833ms)
✔ aggregate limit accepts exact total bytes (5.01725ms)
✔ distribution audit accepts ordinary source and reports no absolute scan path (23.718875ms)
✔ distribution audit flags runtime state without reading its contents (11.113208ms)
✔ distribution audit refuses project media env and archive files (28.261042ms)
✔ home paths and private keys are flagged without echoing secret content (22.862125ms)
✔ malformed text and links fail distribution audit (32.375458ms)
✔ home path scan covers spaces, escaped separators and Windows casing (12.352625ms)
✔ encrypted and legacy private key markers are detected (11.055459ms)
✔ personal preferences and learning files cannot be bundled outside state (1.152291ms)
✔ only an absent state file returns null (1.071708ms)
✔ unreadable or incomplete directory listings cannot imply missing history (0.816ms)
✔ state reader rejects path-shaped names and non-file entries (0.184417ms)
✔ protection documents reject wrong schema malformed lists and duplicate identities (0.165125ms)
✔ corrupt protection state never falls back to a healthy empty registry (1.992834ms)
✔ first-run absence remains healthy and valid protected state remains protected (0.786125ms)
✔ corrupt placement intent cannot be treated as an unused replay-safe request (0.798125ms)
✔ corrupt protection history prevents writes to the registry (0.529625ms)
✔ v5-verify-selects.js rejects unsafe external input before creating state (166.839375ms)
✔ v5-project.js rejects unsafe external input before creating state (147.981667ms)
✔ v5-learn.js rejects unsafe external input before creating state (152.374ms)
✔ bridge reuses exporter with explicit binding and pending receipt (2.0915ms)
✔ replayed request never submits twice (0.43025ms)
✔ initial wrong binding rejected before output creation (0.137917ms)
✔ timeline change at submission blocks host (0.197834ms)
✔ preset changes at submission are refused (0.189291ms)
✔ unknown host outcome blocks replay and records uncertainty (0.176625ms)
✔ pending receipt failure never enters host (0.096041ms)
✔ pins caller request and rejects paths or unsupported range (0.537041ms)
✔ accepted job stays queued when final request-status write fails (0.202417ms)
✔ explicit host rejection is distinguished from uncertain delivery (0.341125ms)
✔ missing underlying APIs never enter submission and remain not_submitted (0.139292ms)
✔ unknown return values remain uncertain and blocked after restart (0.444542ms)
✔ accepted result survives failure of both receipt writes (0.122625ms)
✔ concurrent bridge exports and active edits block host entry (2.21825ms)
✔ matching metadata passes with rational frame rate (0.724084ms)
✔ missing streams, wrong frames, codecs, rate and audio metadata fail (0.232084ms)
✔ invalid expected values fail closed (0.151083ms)
✔ verification decodes both streams with literal argument path and never changes queued receipt (38.758292ms)
✔ missing file and wrong sequence never invoke decoder (20.970292ms)
✔ decoder failure cannot produce successful verification (40.14525ms)
✔ output that changes during decoding is rejected (35.354542ms)
✔ malformed probe output fails without a false pass (7.443958ms)
✔ batch preserves candidate order and summarizes scenes without trimming or ranking (1.046459ms)
✔ duplicate candidate IDs and duplicate aliased source reviews are refused (0.276375ms)
✔ missing source review remains unknown and arbitrary scene IDs cannot alter counters (0.435375ms)
✔ compiled evidence snapshots source identity and segment values (0.08575ms)
✔ indexed review matches the original evaluator across randomized gaps and boundaries (35.349959ms)
✔ malformed and empty batch inputs cannot yield a successful empty report (0.205541ms)
✔ combined review inventory limit applies across source files (0.132792ms)
✔ batch CLI retains scene order and returns nonzero when any candidate needs review (59.599416ms)
✔ prefix review agrees with original for every interval across gaps and focus transitions (3.219334ms)
✔ unreviewed gaps take priority over unknown focus anywhere in the selection (0.113417ms)
✔ prefix boundaries do not count transitions or gaps outside the selected interval (0.059292ms)
✔ empty evidence and outlying candidates stay unreviewed (0.0955ms)
✔ safe integer frame boundaries retain exact rack semantics (0.117166ms)
✔ focus eligibility also requires the actual file digest when verification is requested (2.786708ms)
✔ missing changed and malformed identity results cannot remain eligible (0.483541ms)
✔ soft or unresolved focus candidates do not trigger unnecessary full-file reads (0.159458ms)
✔ batch verifies each eligible source path once and refreshes scene counts (0.574458ms)
✔ input identities are pinned before asynchronous file verification (0.107958ms)
✔ malformed input fails before any file access (0.157625ms)
✔ cancellation does not produce a partial eligible batch report (0.180167ms)
✔ CLI source verification uses real synthetic file content and rejects a changed file (149.170792ms)
✔ file-size refusal is distinct from missing media and cached across candidates (0.532458ms)
✔ file limit validates before IO and cannot be bypassed by an oversized identity result (0.526792ms)
✔ null or string source failures remain unavailable without leaking exceptions (0.1865ms)
✔ focus CLI applies source file limit and refuses limits without verification (183.120083ms)
✔ fully soft reviewed selection is rejected even when rack focus is marked intentional (0.6665ms)
✔ intentional rack into focus is eligible only when selected range includes the sharp finish (0.122666ms)
✔ mixed focus requires review unless it is an explicit rack into focus (0.131041ms)
✔ one missing frame or unknown/unreviewed segment prevents approval (0.383583ms)
✔ sampled frames never establish continuous focus (0.058625ms)
✔ changed source content or fps cannot reuse a focus review (0.057916ms)
✔ exact half-open source frame intervals handle adjacent observations and outer coverage (0.056708ms)
✔ missing evidence is unknown; malformed ranges or provenance are refused (0.150042ms)
✔ focus review does not mutate inputs or select a shot automatically (0.076291ms)
✔ local triple slash aliases are preserved while network identities stay distinct (0.084833ms)
✔ CLI exits successfully only for eligible reviewed focus (130.879459ms)
✔ large adjacent focus inventories remain exact and excessive inventories are refused (21.267459ms)
✔ large placement history uses two directory scans rather than one per receipt (18.224375ms)
✔ bounded inventory scanning still marks duplicate non-file and unreadable evidence malformed (0.384459ms)
✔ receipt publication or removal during history scan prevents a stale complete report (0.315834ms)
✔ valid numeric scalar and host keyframe values remain supported (2.200042ms)
✔ unknown host parameter values cannot masquerade as neutral zero (3.856125ms)
✔ coercible expected and requested grade values never reach transaction (3.036541ms)
✔ coercible target coordinates cannot select a timeline clip (4.118458ms)
✔ NaN and infinite requested values are rejected before mutation (0.729542ms)
✔ successful bridge import becomes the panel import target (0.818875ms)
✔ preflight never replaces current panel import target (0.280875ms)
✔ failed import preserves previous panel import target (0.267166ms)
✔ existing receipt lookup restores panel target after reload (0.856458ms)
✔ missing receipt cannot replace the panel target (0.373875ms)
✔ missing receipt is distinct from a stored null or primitive (4.294458ms)
✔ all known durable receipt states survive reading (1.101625ms)
✔ import receipt adapter rejects duplicate and non-file entries (0.911041ms)
✔ import receipt adapter cannot coerce a non-text file read (0.276833ms)
✔ run pins input before first host await (2.291209ms)
✔ preflight pins input before first host await (0.156792ms)
✔ receipt key remains pinned during storage await (0.160958ms)
✔ restart after lost host response refuses duplicate application (0.341791ms)
✔ completion write failure leaves blocking pending receipt after restart (0.2315ms)
✔ pending persistence failure never enters host (0.095042ms)
✔ post-import inventory failure is uncertain and blocks replay (0.181166ms)
✔ invalid request leaves import lock available (0.967583ms)
✔ malformed sequence inventory never reaches the importer (0.233541ms)
✔ coercible request IDs and filenames are rejected before reading host state (0.435834ms)
✔ edit starting during preflight prevents import and persists not_imported (0.141042ms)
✔ silent pending write loss never enters the importer (0.100292ms)
✔ corrupted pending readback never enters the importer (0.12525ms)
✔ silent completion loss leaves evidence that blocks another import (0.222958ms)
✔ import requires review when a previous sequence disappears or changes (0.504709ms)
✔ project switch after host acceptance requires review (0.094791ms)
✔ successful import retains explicit project and existing-sequence checks (0.093875ms)
✔ project switch during last sequence read prevents host import (0.08375ms)
✔ replaced staged entry with identical XML is rejected before import (0.0785ms)
✔ bin creation failure never imports into the project root (0.075875ms)
✔ host placing sequences outside the work bin requires review (0.065625ms)
✔ stages immutable UTF-8 copy and complete request with no temporary leftovers (15.542708ms)
✔ invalid arguments and XML never create staging state (2.679167ms)
✔ oversized and non-file inputs fail before contents are read (5.180666ms)
✔ invalid UTF-8 is not silently replaced when staging media paths (5.979334ms)
✔ source changing during a bounded read is rejected (1.360334ms)
✔ failed request publication leaves no partial JSON and retry uses a new ID (24.30975ms)
✔ eight simultaneous stagers publish independent complete pairs (173.885916ms)
✔ FIFO inputs are rejected without waiting for a writer (42.0995ms)
✔ 10 MB boundary is measured in bytes and stages the complete document (107.52825ms)
✔ redirected state or imports folder cannot receive staged files (24.044708ms)
✔ dangling state links are not followed or replaced (14.705167ms)
✔ source path replacement during read is rejected (5.691792ms)
✔ linked XML inputs are rejected without creating staging state (14.731541ms)
✔ portable install check passes intact release without exposing absolute paths (20.598167ms)
✔ changed missing and unexpected files fail independently (26.281417ms)
✔ runtime state is never inspected or included in the code manifest (15.403667ms)
✔ invalid manifests traversal duplicates and unexpected symlinks fail closed (5.183208ms)
✔ old Node and manifest symlinks do not receive a healthy setup result (2.348084ms)
✔ Finder metadata does not turn a verified code install into a failure (9.291792ms)
✔ linked install roots cannot redirect an integrity check (1.634041ms)
✔ invalid UTF8 integrity manifests do not receive a healthy result (2.419125ms)
✔ matching but invalid release versions cannot make an installation healthy (295.546708ms)
✔ malformed runtime labels cannot pass a numeric prefix check (30.59875ms)
✔ nested JSON beyond the configured depth is rejected before parsing (0.782083ms)
✔ quotes and escapes cannot spoof structural counts (0.375917ms)
✔ container count is shared across siblings and exact boundaries pass (0.41875ms)
✔ character and option limits reject invalid values without coercion (0.267166ms)
✔ syntax errors never echo private source text (0.228333ms)
✔ ordinary JSON semantics survive bounded scanning (0.112084ms)
✔ backend file reader rejects deep input and still closes its descriptor (2.13625ms)
✔ host state and mask adapters reject excessive nesting rather than treating it as missing (1.049959ms)
✔ host parser errors omit raw JSON content (0.424708ms)
✔ canonical serialization rejects excessive nesting without stack overflow (3.281083ms)
✔ canonical character accounting includes escaped keys and separators exactly (0.179ms)
✔ canonical node limit counts repeated shared objects each time (1.09475ms)
✔ canonical limits refuse oversized strings before invoking JSON.stringify (0.054542ms)
✔ canonical option validation preserves deterministic hashes for ordinary records (0.380292ms)
✔ invalid JSON errors omit snippets of the input document (6.233833ms)
✔ CLI parse failures do not print private input snippets (136.603084ms)
✔ valid JSON retains exact original text and values (33.034917ms)
✔ JSON reader accepts unicode exact-limit input and rejects excess bytes (44.457875ms)
✔ invalid UTF8 symlinks and directories cannot supply command payloads (7.123208ms)
✔ mutation during read is rejected rather than publishing inconsistent input (2.17225ms)
✔ JSON reader enforces explicit bounded limits without relaxing the default (16.947667ms)
✔ document reader returns matching text and parsed value from the same read (47.046208ms)
✔ read-only diagnostic returns bound capture without baseline or mutation dependencies (0.6015ms)
✔ timeline edits between scans refuse learning capture (0.230333ms)
✔ incomplete scans are not accepted as stable snapshots (0.685792ms)
✔ active sequence change during final scan is refused (0.149833ms)
✔ bridge dispatch exposes diagnostic without entering import or export (0.897333ms)
✔ explicit incomplete or malformed completeness flags cannot prove a stable learning scan (0.219375ms)
✔ an incomplete final scan rejects the captured source data (0.107875ms)
✔ absurd host counts fail before enumerating tracks (0.831375ms)
✔ track budget is shared between audio and video (1.478083ms)
✔ item budget is shared between tracks and accepts its exact boundary (0.310833ms)
✔ oversized item collection stops before reading any item metadata (0.213417ms)
✔ invalid capture bounds fail before accessing Premiere (0.171791ms)
✔ unexpected learning CLI arguments are rejected before creating history (133.727458ms)
✔ explicit test purpose remains excluded from editorial profile (139.309417ms)
✔ default comparison purpose remains editorial and is displayed (58.794459ms)
✔ older default-purpose records remain idempotent after the CLI upgrade (145.146708ms)
✔ aggregate byte limit rejects many individually small records (14.748167ms)
✔ record limit refuses a partial history result (0.889667ms)
✔ unrelated and temporary directory entries consume the scan budget (1.875542ms)
✔ invalid limits cannot raise shipped caps or bypass checking absent history (0.388208ms)
✔ history is sorted, malformed records are not skipped, and links are rejected (8.346375ms)
✔ over-limit report leaves existing dashboard and profile intact (87.472209ms)
✔ sparse or malformed edits never become a complete empty comparison (1.120833ms)
✔ unknown completeness flags are not treated as complete (0.337208ms)
✔ source names that collide with object properties retain accurate counts (0.282917ms)
✔ malformed host collections are rejected rather than reported as empty timelines (0.809458ms)
✔ genuine empty track arrays remain supported (0.096166ms)
✔ missing project or sequence identity prevents a learnable capture (0.088084ms)
✔ unreadable item in a valid collection marks capture incomplete (0.118917ms)
✔ ten thousand independent sources preserve exact corrections and totals (42.908625ms)
✔ representable median does not overflow when averaging identical large durations (0.591584ms)
✔ aggregate source durations cannot silently serialize Infinity as null (0.365125ms)
✔ profile duration overflow is refused rather than returning a NaN retention rate (0.340917ms)
✔ small positive median and ordinary timeline measurements remain finite (1.522333ms)
✔ Premiere triple-leading-slash paths match the same absolute source (2.179458ms)
✔ network paths, case, folders and encoded filenames remain distinct (0.638625ms)
✔ path aliases do not inflate unique sources or comparison evidence (0.236042ms)
✔ correction reports preserve the supplied source spelling (0.066625ms)
✔ trim, reject and add durations (0.990625ms)
✔ splits and reorder preserve source retention (0.140666ms)
✔ duplicate occurrences cannot inflate retention (0.109416ms)
✔ added repeat is additional footage use (0.079667ms)
✔ audio and video separate (0.098625ms)
✔ no baseline produces null rather than perfect score (0.072917ms)
✔ malformed, ambiguous, retimed, disabled, incomplete input blocked (0.875291ms)
✔ equal basenames at different paths do not match (0.140792ms)
✔ profile deduplicates and never learns visual taste (0.23875ms)
✔ reference analysis is structural (0.121958ms)
✔ visual metadata has no fabricated measurements (0.360458ms)
✔ 200 interval cases agree with independent integer sampling oracle (10.969792ms)
✔ export basic host clip (0.506333ms)
✔ export unknown capabilities yields incomplete and learning refuses (0.11ms)
✔ export reversed and retimed clips blocked (0.16625ms)
✔ export track enumeration failure propagates (0.081791ms)
✔ export no active project is clear error (0.113ms)
✔ disk records idempotent, profile regenerates, report escapes sequence text (107.239542ms)
✔ panel captures immutable baseline and persists actual comparison (1.5065ms)
✔ panel refuses different active sequence (0.505083ms)
✔ panel refuses incomplete export without recording comparison (0.412292ms)
✔ visual evidence requires provenance and never updates taste implicitly (0.203666ms)
✔ host 2x speed rejected even when in/out duration equals timeline duration (0.089ms)
✔ missing speed capability fails closed (0.369375ms)
✔ explicit nonunit speed refused in supplied decisions (0.045375ms)
✔ current timeline baseline works without loaded decisions (0.865417ms)
✔ unsupported timeline baseline preserves previous baseline (0.51ms)
✔ host numeric boolean flags are accepted, unknown values refused (0.294916ms)
✔ normal-speed mismatches preserve exact host ranges and disclose discrepancy (0.166708ms)
✔ baseline snapshots decisions before the first asynchronous host read (0.377291ms)
✔ baseline ignores replacement decisions during folder resolution (0.309625ms)
✔ sequence switch during export marks the result incomplete (0.163709ms)
✔ closed project during export marks the result incomplete (0.081375ms)
✔ project switch during export cannot be silently accepted (0.075291ms)
✔ sequence-switch failure preserves the previous saved baseline (0.337583ms)
✔ panel reports audio corrections even when video is unchanged (0.709792ms)
✔ generated baseline with conflicting project or sequence identity preserves previous baseline (0.856583ms)
✔ generated baseline checks the active sequence after asynchronous folder lookup (0.812542ms)
✔ timeline baseline checks identity again after asynchronous folder lookup (0.330709ms)
✔ inconsistent saved baseline is rejected before recording a learned comparison (0.266625ms)
✔ matching explicit generated identities are supported (0.351375ms)
✔ panel learning record collisions never overwrite existing evidence (0.439209ms)
✔ ambiguous baseline entries cannot produce a learning record (0.286875ms)
✔ legacy learning refuses linked records before generating a report (76.567583ms)
✔ broken legacy learning root is not treated as empty history (14.11725ms)
✔ legacy learning does not overwrite an existing corrupt comparison (20.289833ms)
✔ legacy learning publication cannot redirect through a linked root (6.998375ms)
✔ dashboard publication does not write through an existing output symlink (77.885ms)
✔ learning outputs use private file permissions and duplicate comparisons stay idempotent (70.775416ms)
✔ learning upgrades recompute derived comparisons without duplicating or rewriting original evidence (53.839375ms)
✔ changed original evidence still blocks a duplicate publication after cache compatibility (41.973958ms)
✔ rectangle bounds reject coercion, zero area, inversions and nonfinite coordinates (1.504125ms)
✔ rectangle application preserves asymmetric edge placement and consumes capture (1.259084ms)
✔ mask write failure cleans up its own effect (0.794208ms)
✔ mask readback failure cleans up its own effect (0.253542ms)
✔ mask append failure cleans up its own effect (0.208083ms)
✔ mask guard failure makes no host transaction (0.107125ms)
✔ mask budget failure makes no host transaction (0.103125ms)
✔ changed timeline and deselection prevent mask application (0.185042ms)
✔ existing Crop and locked tracks are not modified (0.119792ms)
✔ canvas uses its displayed dimensions and drag positions round trip to application (0.821791ms)
✔ host keyframe getters unwrap inherited value properties (0.07025ms)
✔ sparse and prototype-filled coordinates are rejected before host entry (0.124833ms)
✔ capture serializes concurrent loads and rejects application until ready (0.129834ms)
✔ failed capture releases its lock and discards the earlier target (0.072625ms)
✔ mask request is pinned before awaiting host work (0.171792ms)
✔ protection added during preparation is checked again before any mutation (0.098125ms)
✔ malformed track counts stop selection enumeration (0.140875ms)
✔ unreadable selection and lock state fail closed (0.4505ms)
✔ missing identities, inventories and component counts fail closed (0.217541ms)
✔ intent persistence failure stops before Premiere mutation (0.120083ms)
✔ unresolved mask history blocks capture (0.057375ms)
✔ receipt failure after verified application preserves effect and reports uncertainty (0.177667ms)
✔ change during intent write closes a no-mutation outcome (0.119125ms)
✔ host without track-lock API reports the limitation instead of inventing verified status (0.426667ms)
✔ recovery closes verified original no-effect target without host transactions (1.244125ms)
✔ recovery refuses existing Crop, stale selection and changed timeline (0.536083ms)
✔ recovery review cannot be reused after target or intent changes (0.574ms)
✔ failed reconciliation persistence preserves pending status and requires new review (0.421542ms)
✔ mask capture refuses incomplete or warning-bearing timeline inventories (0.159792ms)
✔ project switch during mask capture is detected before returning a capture (0.088583ms)
✔ project switch after append prevents coordinate writes and cross-project cleanup (0.197542ms)
✔ new protection after append prevents both parameter mutation and cleanup (0.140208ms)
✔ source or selection changes after append stop further mutations for review (0.355458ms)
✔ track locked after append stops coordinate writes and automatic cleanup (0.158541ms)
✔ track locking during final component-count read prevents append (0.143042ms)
✔ unfinished mask intent survives journal recreation and prevents replay (1.356959ms)
✔ verified terminal outcomes close exactly their original intent (0.555375ms)
✔ corrupt, orphan and mismatched history is never treated as clean (0.7195ms)
✔ failed terminal persistence leaves a pending intent after restart (0.112875ms)
✔ no-effect reconciliation requires exact binding and retains immutable evidence (0.1625ms)
✔ terminal receipts cannot close forged or changed intents (0.197875ms)
✔ reconciliation without matching stored evidence is corrupt, not completed (0.23125ms)
✔ journal survives an actual child-process exit using persisted intent files (61.575334ms)
✔ overlapping journal writers cannot publish two intents in the same sequence (0.222083ms)
✔ completed mask history detects changed intent coordinates (0.24725ms)
✔ legacy receipts remain readable and are explicitly counted as weaker evidence (0.105042ms)
✔ new receipts carry complete intent evidence and no legacy count (0.169834ms)
✔ journal writer lock releases after storage refusal (0.128916ms)
✔ repeated concurrent calls produce at most one intent per journal instance (4.193625ms)
✔ terminal request remains pinned while journal reads are pending (0.447209ms)
✔ partial receipt on disk is preserved and never silently repaired (2.494167ms)
✔ mask journal rejects invalid history limits before using storage (0.157375ms)
✔ combined record limit applies before reading either history folder (0.123ms)
✔ aggregate characters include terminal receipts and accept the exact boundary (0.1285ms)
✔ new mask operation requires room for both intent and completion (0.093875ms)
✔ character capacity refusal leaves no new intent (0.043917ms)
✔ pending operations in other sequences retain reserved receipt capacity (0.51175ms)
✔ reserved character capacity fits no-effect reconciliation exactly (0.104ms)
✔ a large binding cannot start an intent whose receipt exceeds storage limit (0.419916ms)
✔ exact minimum rectangle size remains valid away from the origin (0.693667ms)
✔ blank and nondecimal fields cannot silently become valid mask coordinates (1.127541ms)
✔ invalid field can be corrected without losing the captured clip (0.202292ms)
✔ coordinate edits cannot reenable Apply during a pending host operation (0.182833ms)
✔ unavailable drawing surface is reported without an unhandled event exception (0.211083ms)
✔ Apply validates current field text even before a change event is delivered (0.107792ms)
✔ storage writes once, reopens persisted entry and verifies exact data (1.109791ms)
✔ storage rejects invalid folder and traversal before requesting access (0.583667ms)
✔ storage refuses scalar, malformed, empty and oversized records (0.487583ms)
✔ storage rejects unsafe inventory shapes and duplicate names (0.183166ms)
✔ storage refuses wrong entry identity even with valid JSON (0.108958ms)
✔ storage write readback detects truncated or replaced files (0.173875ms)
✔ storage write snapshots caller data before folder access yields (0.097459ms)
✔ storage refuses oversized outgoing records before creating a file (0.207208ms)
✔ permission loss after write reports uncertainty without removing the record (0.125167ms)
✔ missing filenames are rejected without opening an entry (0.107791ms)
✔ isolated storage probe verifies the real adapter contract (0.143958ms)
✔ rectangle matte isolates center and leaves surroundings unchanged (1.419583ms)
✔ ellipse and rotation respect frame aspect ratio (0.214417ms)
✔ polygon winding does not change fill and concave shapes work (1.50275ms)
✔ feather transitions smoothly and expansion grows selection (1.045458ms)
✔ inversion is complementary and opacity multiplies final coverage (0.420542ms)
✔ add, subtract and intersect follow documented matte algebra (0.452417ms)
✔ zero opacity and empty corrections are identity; sequential layers use previous output (0.669542ms)
✔ self intersections, degenerate shapes, tracking and invalid numerics fail closed (0.244792ms)
✔ malformed images, oversize rasters and unsupported cube masks rejected (0.153959ms)
✔ color plan preserves masks and labels LUT as base-only (0.299583ms)
✔ CLI exports correctly dimensioned binary matte with manifest binding (22.754292ms)
✔ streamed source hashes match independent hashes across chunk boundaries (121.693625ms)
✔ matching and mismatching expected digests are explicit (18.142333ms)
✔ links and directories are not hashed (20.140917ms)
✔ invalid options fail before touching a missing file (0.2455ms)
✔ growth shrinkage and same-size replacement during reading invalidate identity (47.610042ms)
✔ in-place same-size mutation is detected even after its bytes were read (6.485292ms)
✔ cancellation before and during hashing never reports a verified digest (9.453ms)
✔ CLI returns distinct success mismatch and error codes without source paths (119.56ms)
✔ file handles close after progress callback failures and deleted-path failures (49.35ms)
✔ progress is monotonic and bounded by the requested read chunk (11.944375ms)
✔ source byte cap rejects oversized files before opening their contents (18.760834ms)
✔ zero byte cap accepts empty files and validates options before source access (8.111625ms)
✔ identity CLI byte cap is strict and composes with expected digest (199.441291ms)
✔ apply_warp_stabilizer_last_video cannot bypass sequence protection (1.449917ms)
✔ apply_warp_stabilizer_target cannot bypass sequence protection (0.311167ms)
✔ all currently exposed timeline mutation commands pass the protection gate (3.580375ms)
✔ read-only session status remains available without mutation authorization (0.245083ms)
✔ active import blocks stabilizer before the protection or mutation call (0.501125ms)
✔ actual protection guard refuses protected sequences and unhealthy history (1.589458ms)
✔ all learning style and generation actions enter the shared gate before host reads (1.392042ms)
✔ changing timeline cannot publish a mixed learning baseline (0.724292ms)
✔ stable panel baseline records the verified timeline token (0.469084ms)
✔ empty UXP dropdowns receive valid visible defaults (0.622458ms)
✔ mounting controls preserves a valid explicit choice (0.088041ms)
✔ missing optional controls do not prevent panel mounting (0.083833ms)
✔ empty pilot results stay incomplete with no invented timings (0.682916ms)
✔ reported pilot gates never turn into public-release approval or verified evidence (0.745709ms)
✔ repeating a machine or tester cannot inflate independent-installation counts (0.233667ms)
✔ every installation lifecycle check is required, including uninstall and rollback (0.293042ms)
✔ picture/audio review and recovery cannot be replaced by structural checks (0.820041ms)
✔ one critical incident pauses pilot expansion even with enough passing sessions (0.413417ms)
✔ earlier versions do not count toward current release gates or timing (0.242833ms)
✔ incomplete timing is omitted and negative savings are reported honestly (0.133166ms)
✔ duplicate records, invalid counts and unexpected private-data fields are rejected (0.422583ms)
✔ missing outcomes, nonfinite durations and invalid inventories fail explicitly (0.304875ms)
✔ three tester IDs and machine IDs require three nonoverlapping installation pairs (0.179208ms)
✔ bumping the target version does not silently dismiss a recorded incident (0.09325ms)
✔ state folder type confusion cannot imply absent replay history (0.654875ms)
✔ malformed receipt cannot claim applied (3.346208ms)
✔ valid placement statuses preserve unused uncertain and applied distinctions (1.101666ms)
✔ malformed history blocks new placement advisory even without a request ID (2.152875ms)
✔ valid completed history stays healthy and unresolved current scope blocks (0.741584ms)
✔ history rejects non-file JSON entries and unreadable listings (0.179292ms)
✔ actual receipt writer produces records accepted by status reader (1.496542ms)
✔ alternate encoded filename cannot create duplicate logical request history (5.595ms)
✔ isolates projects with the same sequence and media (83.399875ms)
✔ canonical deduplication survives object key order and process restart (33.775125ms)
✔ rejects sequence and project mismatches before writing (9.346667ms)
✔ rejects incomplete and invalid numeric documents (4.964334ms)
✔ tampering fails audit and blocks partial reports and global learning (36.412291ms)
✔ truncated records surface errors and abandoned temporary files are ignored (42.206917ms)
✔ path-like project IDs cannot escape the store (13.355958ms)
✔ fresh is default, remember persists, switching back preserves learning (52.635917ms)
✔ corrupt style configuration fails closed (44.029ms)
✔ CLI imports panel evidence, ignores cached scores, and returns failure status on corruption (273.890291ms)
✔ concurrent imports publish exactly one complete immutable record (119.507709ms)
✔ protecting more than 200 sequences retains their latest protection evidence (38.436667ms)
✔ same-millisecond protection changes do not reuse an event ID (0.607292ms)
✔ frequent changes on one sequence cannot erase a different protected sequence (22.998541ms)
✔ preexisting inconsistent protection evidence is not silently repaired or overwritten (0.636ms)
✔ interrupted two-file publication remains visible and blocks mutations (0.504ms)
✔ concurrent publication cleanup cannot invalidate an identical existing record (65.089791ms)
✔ published-record reader does not retry content changes (25.875167ms)
✔ link metadata retries remain bounded and external input reads stay strict (17.635833ms)
✔ actual queue adapter persists intent before host entry and consumes confirmed result (2.132834ms)
✔ actual queue adapter retains request during result outage and never reexecutes after restart (2.585875ms)
✔ corrupt persisted intent remains queued without overwriting evidence or entering host (0.498625ms)
✔ sender refuses a linked commands folder without publishing outside state (52.779208ms)
✔ sender refuses a linked state root without publishing into its target (37.196167ms)
✔ invalid payload files do not create an empty command queue (32.51875ms)
✔ valid read-only command creates one private command in a new queue (68.863292ms)
✔ sender refuses dangling queue links and non-directory roots (98.693ms)
✔ sender rejects malformed JSON and mask IDs without state changes (97.884333ms)
✔ queued identity confusion is rejected before any command handler runs (3.9245ms)
✔ valid queued command reaches dispatcher once and releases polling lock (0.441584ms)
✔ invalid JSON is consumed without invoking a handler or stranding the queue (0.390459ms)
✔ filename boundaries and command identifiers are strict (0.157625ms)
✔ unreadable queued file remains available for retry (0.415084ms)
✔ delivery persistence failure retains the queued request (0.478042ms)
✔ recovery rejects another project even when sequence and timeline tokens match (1.467709ms)
✔ recovery refuses absent or conflicting binding evidence before operation preflight (1.991541ms)
✔ matching recovery binding reaches normal preflight without performing an edit (0.452708ms)
✔ changed timeline or previous recovery still blocks valid receipt (0.90875ms)
✔ truthy recovery safety flags cannot authorize automatic placement (3.214333ms)
✔ malformed warning and placement inventories cannot disappear through defaults (1.043792ms)
✔ missing original target is detected before recovery placement preflight (0.347208ms)
✔ empty or malformed placement edits cannot become a successful empty recovery (1.294792ms)
✔ valid single-item recovery still reaches normal placement preflight (0.360667ms)
✔ recovery placement must match the original source range position and track (1.217458ms)
✔ ripple linked-media retimed reversed and disabled targets require manual review (1.489958ms)
✔ normal-speed audio-only source recovery remains supported (0.320166ms)
✔ placement recovery requires an original non-shifting removal (1.405375ms)
✔ bundle extracts into an intact companion and a matching panel archive (125.202ms)
✔ identical release inputs create byte-identical complete bundles (66.117917ms)
✔ bundle refuses existing output and leaves it unchanged (90.050292ms)
✔ private runtime state and modified source cannot enter a bundle (71.976042ms)
✔ bundle rejects output inside release or redirected output parents (146.153125ms)
✔ generated profiles never invent an editor identity (0.7345ms)
✔ current README version and historical-document pointers are consistent (0.203541ms)
✔ release source passes the distribution file and privacy audit (24.078084ms)
✔ latest uses write time across differently prefixed command IDs (41.462291ms)
✔ specific ID and filename ignore ordering and summary preserves failure (65.5205ms)
✔ unknown flags extra positionals invalid waits and paths fail explicitly (163.633125ms)
✔ corrupt newest result is not hidden by falling back to older success (46.53475ms)
✔ mismatched envelope identity is refused (66.227458ms)
✔ wait observes delayed complete publication (204.1145ms)
✔ symlink result cannot read an unrelated target (57.816041ms)
✔ FIFO result does not hang (42.300666ms)
✔ malformed UTF8 result cannot be reported as successful (31.981625ms)
✔ linked result directory cannot redirect result inspection (26.021292ms)
✔ symlink replacement after result stat is rejected (33.790709ms)
✔ result reader preserves decoded JSON formatting without a second file read (31.642875ms)
✔ partial result publication can complete during the bounded wait (130.788916ms)
✔ oversized result is rejected by its size before parsing (27.075208ms)
✔ broken results-directory link reports an error instead of missing history (25.104416ms)
✔ linked state root cannot redirect result inspection (26.192792ms)
✔ dangling state root is corruption rather than absent history (24.8965ms)
✔ summary keeps uncertain delivery distinct from ordinary command failure (28.011292ms)
✔ known result lookup does not enumerate unrelated history (23.452417ms)
✔ latest refuses an oversized inventory without returning partial success (23.909209ms)
✔ result wait deadline does not depend on advancing wall-clock time (59.250542ms)
✔ result waits require unambiguous integer milliseconds (156.191875ms)
✔ editor benchmark separates shared missed and rejected source ranges (1.066125ms)
✔ paths require explicit opt-in and input labels cannot leak through report (0.761458ms)
✔ reordering preserves overlap and repeated occurrences cannot inflate shared duration (0.183625ms)
✔ comparison CLI reads without creating learning history and rejects unknown options (70.188208ms)
✔ retimed or incomplete documents cannot generate benchmark scores (0.166041ms)
✔ legacy requests receive balanced guidance without claiming a performed review (1.405208ms)
✔ body language guidance survives request serialization and reaches generation context (0.151167ms)
✔ invalid priorities are rejected rather than silently converted to balanced (0.132416ms)
✔ guidance instances cannot alter another request (0.0715ms)
✔ capture publishes the selected priority bound to the active project (0.139291ms)
✔ invalid capture priority cannot reach the host or publisher (0.612542ms)
✔ priorities leave selection of behavior to the current brief (0.108416ms)
✔ panel export includes the visible priority and restores controls after failure (1.082333ms)
✔ read-only first run reports missing state without creating files (24.135083ms)
✔ explicit initialization creates empty queues without claiming a host connection (47.399583ms)
✔ repeat initialization preserves queued commands and recovery history byte-for-byte (37.468916ms)
✔ unverified release and unsupported runtime cannot create state (46.278666ms)
✔ existing incomplete folders are left untouched (62.948416ms)
✔ redirected root, queue and parent directories are rejected (54.6735ms)
✔ partial initialization failure is reported and preserved for inspection (103.811708ms)
✔ setup CLI rejects ambiguous and repeated options (3.758209ms)
✔ successful stabilization reserves rollback and counts both host transactions (1.567041ms)
✔ insufficient rollback capacity refuses all host mutation (1.589042ms)
✔ failed set counts attempted transaction and successful removal (0.497416ms)
✔ failed readback counts both forward transactions plus rollback (0.347084ms)
✔ rejected append consumes one conservative attempt without adding an effect (0.305667ms)
✔ unknown set outcome cannot disappear from the session count (0.298708ms)
✔ 37 used units can complete rollback without exceeding 40 (0.355292ms)
✔ reserved path components never reach state read or folder creation (0.678458ms)
✔ newly created folder must have the requested name and folder type (0.177292ms)
✔ state reads cannot coerce a non-text host result into valid JSON (0.123542ms)
✔ ordinary folder names and Unicode JSON still work (0.409708ms)
✔ default history remains under the existing bridge state location (0.391958ms)
✔ custom bridge root is shared by learning and project tools (154.6515ms)
✔ explicit legacy overrides remain independent and do not migrate history (9.614209ms)
✔ legacy project override wins in the actual CLI (49.896542ms)
✔ legacy learning override wins in the actual CLI (88.522917ms)
✔ canonical records do not execute getters or serialization hooks (0.559709ms)
✔ hidden properties and symbols cannot disappear from stored evidence (0.144542ms)
✔ custom array prototypes are not canonical plain JSON records (0.065417ms)
✔ frozen plain JSON remains valid with stable hashes (0.293667ms)
✔ dynamic document values cannot publish an unreadable record (3.437958ms)
✔ sparse arrays cannot collapse to empty arrays or generate invalid JSON (0.718125ms)
✔ circular records fail explicitly while shared noncyclic values stay valid (0.116291ms)
✔ sparse record metadata is rejected before publishing history (0.943375ms)
✔ finite JSON canonicalization preserves content and key-order identity (2.29475ms)
✔ global profile reads one validated record set per project (45.755625ms)
✔ global profile rejects errors in its only audited record set (29.496916ms)
✔ global profile rejects symbolic-link records before reading content (60.774333ms)
✔ 300 deterministic JSON records preserve canonical round trips (20.859042ms)
✔ multi-project history preserves counts after restart and excludes test evidence (828.807375ms)
✔ dangling style settings link is not silently treated as fresh (42.08125ms)
✔ style settings symlinks cannot substitute a valid settings file (38.811875ms)
✔ redirected project history is refused by local audit (52.024458ms)
✔ project publication refuses a redirected directory before writing outside state (26.840125ms)
✔ style publication refuses redirected settings directory (24.024459ms)
✔ invalid UTF8 cannot validate by replacement-character decoding (79.823209ms)
✔ record replacement with a symlink during validation is refused (120.156583ms)
✔ genuinely absent project history and settings retain their normal defaults (26.351875ms)
✔ dangling project history link reports uncertainty instead of empty history (20.715375ms)
✔ linked store root cannot redirect comparison publication (34.374416ms)
✔ linked store root cannot redirect style publication (6.078917ms)
✔ linked store root cannot supply foreign history or style (52.753708ms)
✔ dangling store root is not mistaken for fresh or empty history (5.774667ms)
✔ killed writer before publication recovers without partial history (115.163542ms)
✔ killed writer after publication recovers without partial history (116.9995ms)
✔ 12 independent writers publish exactly one immutable record (165.615917ms)
✔ project audit enforces the exact aggregate input-byte boundary (55.313292ms)
✔ multiple individually valid records cannot bypass the byte limit (37.174625ms)
✔ global profiles share the record limit across projects (70.994125ms)
✔ global discovery and audit reads share the byte budget (78.304292ms)
✔ invalid history scan limits are rejected before creating a store (0.735833ms)
✔ scan limits do not delete history or prevent bounded known-record inspection (80.587834ms)
✔ directory entry limits include ignored temporary files and fail without a partial report (42.289083ms)
✔ global directory inventory shares its entry budget across projects and discovery (78.563542ms)
✔ directory handles close when inventory limits interrupt a scan (25.864334ms)
✔ defaults fresh and persists independent project choices across controllers (0.68325ms)
✔ switching projects before save refuses to write (0.205625ms)
✔ unsaved project and missing load refuse writes (0.448375ms)
✔ invalid choice and corrupted preferences fail closed (0.118375ms)
✔ publication failure is explicit, recoverable, and does not claim sync (0.165875ms)
✔ concurrent save is blocked while publication is pending (2.135792ms)
✔ duplicate project entries and malformed modes rejected (0.130958ms)
✔ panel artifact reaches actual backend CLI and fresh overrides existing remembered mode (204.139791ms)
✔ panel buttons load, save, export and surface active-project mismatch (1.050083ms)
✔ silent preference write loss cannot publish a saved backend choice (0.132375ms)
✔ changed preference readback prevents backend synchronization (0.157916ms)
✔ matte parser preserves whitespace-valued first pixels and catches truncation (0.969125ms)
✔ checksum, frame identity, review state and dimensions prevent wrong-frame correction (0.598417ms)
✔ subject and background corrections affect complementary regions without changing source (1.708792ms)
✔ feather and zero opacity work (0.807ms)
✔ translation tracks a textured subject and marks result unreviewed (1.843666ms)
✔ textureless subjects and scene cuts return lost without a mask (1.174125ms)
✔ tracking rejects frame gaps, reverse playback and mismatched media (0.228583ms)
✔ report includes install and runtime status without creating missing state (50.279792ms)
✔ filenames payloads IDs and error contents never enter the support report (108.444916ms)
✔ unreadable redirected delivery inventory is summarized without leaking its target (77.228584ms)
✔ untrusted version and runtime labels cannot smuggle arbitrary content (65.697542ms)
✔ CLI writes a private report exclusively and preserves an existing destination (184.961542ms)
✔ empty initialized state is distinguishable from incomplete setup (16.690958ms)
✔ linked results folder cannot produce a healthy support inventory (11.311625ms)
✔ test comparisons never enter learned retention even when perfect (2.561417ms)
✔ test-only history returns no learning signal (0.087625ms)
✔ unknown purpose cannot silently enter profile (0.268583ms)
✔ project import and global profile preserve exclusion and duplicate identity (81.321917ms)
✔ HTML report exposes purpose and timing discrepancies (78.008459ms)
✔ learning buttons share a lock and release it after failure (4.107ms)
✔ malformed clip collections produce warnings rather than trustworthy empty scans (3.986875ms)
✔ invalid track counts cannot produce a valid inventory token (2.119ms)
✔ real empty tracks are distinct from failed enumeration (0.374667ms)
✔ unknown host time and boolean values remain unknown (0.249709ms)
✔ timeline tokens distinguish unreadable values from zero and false (0.711167ms)
✔ unknown mute state is not fabricated as false (2.689333ms)
✔ actual clip description preserves missing times, speed and flags (0.492167ms)
✔ timeline inventory refuses extreme and combined track counts before reading tracks (0.653125ms)
✔ oversized clip arrays do not become warning-only partial snapshots (0.363708ms)
✔ timeline item budget spans audio and video and accepts its exact boundary (0.4595ms)
✔ warning-bearing inventories explicitly report incomplete status (0.269792ms)
✔ invalid timeline bounds fail before host reads (0.273667ms)
✔ missing state never counts as a clear queue and is not created (100.365125ms)
✔ complete empty local checks require separate host review (105.294333ms)
✔ queued status requests block switching even though read-only (21.693625ms)
✔ recorded requests still queued are not silently cleared (33.60925ms)
✔ unfinished intent remains uncertain and private details are omitted (13.932792ms)
✔ corrupt delivery evidence cannot pass the preflight (5.550292ms)
✔ symlinked state and incomplete queues require attention (7.332084ms)
✔ failed installation stays blocked with a clear queue (12.419459ms)
✔ CLI success returns 0 and preserves input files (90.102666ms)
✔ CLI mismatch returns 1 and preserves input files (52.98925ms)
✔ CLI invalid input returns 2 and preserves input files (48.801583ms)
✔ CLI missing arguments returns usage error (32.877083ms)
✔ CLI missing files returns read error (29.016542ms)
✔ one source frame slip cannot hide inside the seconds tolerance (1.110667ms)
✔ one timeline frame slip cannot hide inside the seconds tolerance (0.1505ms)
✔ high-rate manifest retiming is refused even below the old seconds epsilon (0.210666ms)
✔ high-rate timeline overlap is refused even below the old seconds epsilon (0.156458ms)
✔ common camera rates still accept floating-point noise far below one frame (0.239292ms)
✔ frame ranges that collapse during seconds conversion are refused (0.087166ms)
✔ snapshot scan warnings and incomplete state cannot pass (0.921375ms)
✔ timeline scan warnings and incomplete state cannot pass (0.173875ms)
✔ malformed warnings are rejected (0.19ms)
✔ empty warnings and explicit completeness pass (0.068667ms)
✔ sparse manifest rejected with actionable error (0.064166ms)
✔ 200 seeded missing-item, source collision, duplicate and slip cases (15.921209ms)
✔ sparse snapshot tracks and clip lists cannot hide unreadable entries (0.187625ms)
✔ rejects muted (0.940833ms)
✔ rejects unknown mute (0.123625ms)
✔ rejects track index (0.095459ms)
✔ rejects track count (0.093958ms)
✔ rejects item track (0.103834ms)
✔ rejects total count (0.087875ms)
✔ extreme fps overflow cannot pass (0.179792ms)
✔ 1000 fractional-rate repeated ranges and sparse audio mutations (14.931084ms)
✔ indexed selects verifier matches frozen original across 1000 mutated cases (202.861625ms)
✔ indexed verifier retains exact tolerance behavior at floating point boundaries (0.600125ms)
✔ large repeated-source selects verify without losing per-channel matching (136.126375ms)
✔ frame-aware tolerance distinguishes neighboring starts while rejecting actual duplicates (1.501958ms)
✔ exact video/audio passes without changing input (1.38075ms)
✔ wrong project_guid fails (0.198459ms)
✔ wrong sequence_guid fails (0.106667ms)
✔ wrong timeline_state_token fails (0.094625ms)
✔ one frame source offset fails (0.120292ms)
✔ same filename in another directory fails (0.092917ms)
✔ missing duplicated and misplaced audio channels fail (0.167125ms)
✔ audio slip fails even when picture passes (0.084625ms)
✔ no audio passes when none expected (0.079916ms)
✔ unknown speed and enabled state fail (0.169041ms)
✔ malformed manifests and missing evidence throw (0.211ms)
✔ successful wrappers accepted and failed commands rejected (0.608375ms)
✔ native links outside verified scope (0.247542ms)
✔ network-style double-slash source cannot alias a local absolute path (0.108959ms)
✔ whole sequence queues expected host arguments and receipt, never claims completion (2.137875ms)
✔ In/Out range reaches host and MOV follows preset (0.264625ms)
✔ preset preference separate from proxy preset (0.076958ms)
✔ cancel preset or destination makes no export call (0.128917ms)
✔ reject invalid filename, no project, no sequence, no AME, missing capability (0.336833ms)
✔ changed active sequence blocked after chooser (0.188583ms)
✔ audio/image preset rejected (0.086791ms)
✔ host rejects submission without success receipt (0.092917ms)
✔ unknown or thrown host response warns against blind retry (0.169459ms)
✔ receipt failure still returns accepted job (0.176416ms)
✔ folder collision refuses overwrite (1.5445ms)
✔ stale token recovers through chooser (0.280167ms)
✔ concurrent requests blocked, lock released after cancellation (3.231833ms)
✔ extension removal cannot produce empty or invalid export stem (0.247792ms)
✔ invalid post-extension names never queue export or create folders (0.186916ms)
✔ uses static extension helper and documented queue enum without instance helper (0.075375ms)
✔ observed values cannot disappear or change meaning during JSON serialization (0.909125ms)
✔ nested unsupported values and cyclic evidence values are refused (0.155792ms)
✔ sparse evidence references cannot satisfy provenance requirements (0.449459ms)
✔ inherited evidence references do not survive a JSON round trip (0.114541ms)
✔ finite structured observations survive round trip without inferring preferences (0.604791ms)
✔ observation validation does not invoke getters or serialization hooks (0.115375ms)
✔ deep and oversized observations fail with bounded validation errors (2.296041ms)
✔ creates a new work bin without adopting or moving human items (0.643375ms)
✔ project change prevents bin transaction (0.400958ms)
✔ rejected creation cannot supply an import destination (0.100292ms)
✔ ambiguous new items fail closed instead of choosing a bin (0.1115ms)
✔ sequence containment accepts nested imports and rejects human-root sequences (0.853333ms)
ℹ tests 805
ℹ suites 0
ℹ pass 805
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 1790.472334
