Benchmark

A request is a plain-language automation instruction for a workspace of installed apps. The benchmark has 149 feasible requests, which the available app capabilities can support, and 55 infeasible requests, which require missing capabilities and should be refused.

An oracle state is a seeded set of app databases, events, and clock steps with checks for the requested behavior. An attempt is one independent authoring run, including any repairs allowed by the pipeline; a feasible attempt passes only if its final workflow passes all scored oracle states, and an infeasible attempt passes when it correctly refuses.

Each count below is final passes out of three attempts. The full pipeline includes table definitions (GR); the Luna comparison omits them (gR). “Failed by any model” keeps requests with at least one failed attempt in any of these three conditions. Expand a request to see its oracle state names and the first reported failure for each condition that failed.

Adjudication rule

The rule section of bench/ADJUDICATION.md, quoted verbatim:

R1 — Open wording/content choices: accept reasonable greetings, summaries, identifiers, field selection, and wording that the request leaves unspecified.
R2 — Visible contracts: do not demand a tag encoding, text-link convention, or event-only implementation that exists only in fixture setup.
R3 — Empty reports: accept silence or one truthful zero/none report for an empty digest/count; retain silence for empty per-item or explicit conditional actions.
R4 — Dates and schedules: accept reasonable unspecified dates, calendar-week starts, monthly cohorts, and unqualified schedule phases. Explicit times and named windows remain binding.
R5 — Contradictions: retain checks for wrong destinations/recipients, excluded rows, wrong windows, duplicates, runtime errors, and complete tag/word filters.

Download requests and results (JSON) · Includes all attempts, first and final scores, failure messages, and source hashes. Email addresses are replaced with [address].

204 of 204 requests

204 requests · final passes out of 3
ID	Request	Family	Installed apps	Feasibility	Luna Full pipeline	Sol Full pipeline	Luna No table definitions
aggregate-01	Every Monday at 9 AM, count my unread Work Emails and send me a summary notification. Oracle states and failures Oracle states two unread work inbox messages no unread work inbox messages empty data First failure per condition Luna · without table definitions (attempt 1): refused a feasible request	aggregate	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
aggregate-02	Every Friday at 5 PM, count tasks completed this week and create a summary note. Oracle states and failures Oracle states this-week completions and week-boundary decoys only previous week empty data First failure per condition Luna · without table definitions (attempt 1): [this-week completions and week-boundary decoys] expected 1 deliveries to com.lattice.notes::add-note, got 0	aggregate	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
aggregate-03	Every Monday, send a digest to my Work Email of tasks completed last week, grouped by category. Oracle states and failures Oracle states last week categories and week boundaries nothing completed last week empty data First failure per condition Luna · without table definitions (attempt 1): [last week categories and week boundaries] expected 1 deliveries to com.lattice.email#work::send-email, got 0	aggregate	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
aggregate-04	Every Sunday, count bookmarks saved this week grouped by tag, and email me a reading-list summary from my Personal Email. Oracle states and failures Oracle states individual tags with duplicates and week decoys only bookmarks from prior week empty data different tag counts First failure per condition Luna · without table definitions (attempt 1): [individual tags with duplicates and week decoys] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	aggregate	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
aggregate-05	Every month, aggregate my notes by tag, count entries per tag, and send a summary to my Personal Email. Oracle states and failures Oracle states individual tags across all note ages empty data unequal tag totals First failure per condition Luna · without table definitions (attempt 1): [individual tags across all note ages] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	aggregate	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
aggregate-06	Every Sunday at 6pm, send me an email summarizing all the money spent on Splitwise this week with the grand total. Oracle states and failures Oracle states weekly purchases with separate USD/EUR grand totals empty ledger different decimal total First failure per condition Luna · without table definitions (attempt 1): [weekly purchases with separate USD/EUR grand totals] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	0 / 3
aggregate-07	Every Monday morning at 8am, send a desktop notification with the total distance in kilometers I ran on Strava over the last 7 days. Oracle states and failures Oracle states meters converted, exact seven-day lower boundary no runs fractional kilometer First failure per condition Luna · without table definitions (attempt 1): [meters converted, exact seven-day lower boundary] expected 1 deliveries to com.lattice.notifications::receive-notification, got 0	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	0 / 3
aggregate-08	On the first day of every month at 9am, make a note in Notes listing all the books I finished reading in Goodreads during the last 30 days. Oracle states and failures Oracle states slash dates and unrated finished books no finished books in window empty reading log First failure per condition Luna · without table definitions (attempt 1): [no finished books in window] missing explicit zero/none summary	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	0 / 3
aggregate-09	Every Saturday morning at 10am, send an email to my personal inbox with the count of all unread notification threads in GitHub Inbox. Oracle states and failures Oracle states string IDs and unread flag empty inbox can report zero or stay silent no scheduler tick means no mail First failure per condition Luna · without table definitions (attempt 3): refused a feasible request	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	2 / 3
aggregate-10	On the last day of each month at 6pm, total up all the hours and miles I logged in Strava over the last 30 days and record it in a training log note in Notes. Oracle states and failures Oracle states leap February end, rolling timestamp boundaries January 31 also qualifies January 30 is not month end month end without training data non-leap February month end thirtieth calendar date counts month-end guard must suppress wrong hours First failure per condition Luna · full pipeline (attempt 3): [leap February end, rolling timestamp boundaries] missing 3.5 with hours?\|hrs?\|h unit Luna · without table definitions (attempt 1): [leap February end, rolling timestamp boundaries] expected 1 deliveries to com.lattice.notes::add-note, got 0	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	2 / 3	3 / 3	0 / 3
aggregate-11	Every Sunday at 7pm, count how many books Goodreads says I finished in the past week and email me the total. Oracle states and failures Oracle states slash dates; past-week windows none finished this week one finished First failure per condition Sol · full pipeline (attempt 1): [slash dates; past-week windows] expected count 3 or 4, found [3,7] Luna · without table definitions (attempt 1): [slash dates; past-week windows] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	0 / 3	0 / 3
aggregate-12	Every Monday at 8am, count the new GitHub Inbox threads received in the previous week and send me the number as a notification. Oracle states and failures Oracle states new threads observed during the week; an updated old thread is not new quiet week one thread First failure per condition Sol · full pipeline (attempt 3): [new threads observed during the week; an updated old thread is not new] expected 1 deliveries to com.lattice.notifications::receive-notification, got 0	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	2 / 3	3 / 3
aggregate-13	Every Friday at 4pm, count how many recipes were saved in Recipe Box that week and email me the total. Oracle states Oracle states saves observed this week; publication dates irrelevant nothing saved one save mid-week All three conditions passed every attempt.	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
aggregate-14	On the first day of each month at 10am, count the expenses added to Splitwise in the previous month by currency and email me the breakdown. Oracle states and failures Oracle states previous calendar month by currency settlements may or may not count as expenses no expenses last month First failure per condition Luna · full pipeline (attempt 1): [previous calendar month by currency] expected 1 deliveries to com.lattice.email#personal::send-email, got 0 Luna · without table definitions (attempt 1): [previous calendar month by currency] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	2 / 3	3 / 3	0 / 3
aggregate-15	Every Sunday at 6pm, total the quantities consumed from Grocy this week by product and email me each total with its unit. Oracle states Oracle states observed usage, old backlog and stock balance decoys fractional quantities are summed no consumption All three conditions passed every attempt.	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
aggregate-16	On the first of each month at 8am, count my Goodreads books on each shelf and send me the counts in an email. Oracle states Oracle states three distinct shelf counts single shelf, empty groups optional exclusive and custom shelf readings empty library All three conditions passed every attempt.	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
aggregate-17	Every Monday at 7am, total last week’s Strava distance separately for runs and rides and put both kilometer totals in a note. Oracle states and failures Oracle states per-type previous week kilometer totals calendar week endpoint choices public run and ride subtypes no previous-week activities First failure per condition Luna · without table definitions (attempt 1): [per-type previous week kilometer totals] expected 1 deliveries to com.lattice.notes::add-note, got 0	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	0 / 3
aggregate-18	Every weekday at 8am, count my overdue Todoist tasks by priority and send me the counts as a notification. Oracle states and failures Oracle states unfinished overdue grouped API priorities second priority distribution no overdue tasks First failure per condition Luna · without table definitions (attempt 3): [unfinished overdue grouped API priorities] wrong/missing labeled priority\s4\|api\s4\|urgent: 0 expected 2	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	2 / 3
aggregate-19	At 9pm each Sunday, count the emails I received on each day of the past week and save the daily counts in Notes. Oracle states and failures Oracle states received only, per-date counts second daily counts past-week endpoints and future receipt exclusion no received emails First failure per condition Luna · full pipeline (attempt 2): [received only, per-date counts] wrong/missing labeled 2027-03-02\|2027\/0?3\/0?2\|0?3\/0?2(?:\/2027)?\|0?2\/0?3(?:\/2027)?\|\bmar[a-z]\s+0?2\b\|\b0?2\s+mar[a-z]\b\|\btue(?:sday)?\b: expected 2 Sol · full pipeline (attempt 3): [no received emails] an empty report contains a nonzero count Luna · without table definitions (attempt 1): [received only, per-date counts] expected 1 deliveries to com.lattice.notes::add-note, got 0	aggregate	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	1 / 3	2 / 3	0 / 3
aggregate-20	Every Sunday at 9pm, count my Habits checks and missed periods from the past seven days and append both numbers to today’s Journal entry. Oracle states and failures Oracle states matching checks and missed periods nonmatching old history and successful periods boundary either seven-date or full prior-day window empty history First failure per condition Luna · full pipeline (attempt 1): [matching checks and missed periods] wrong/missing /miss/: expected 1, got 2 Luna · without table definitions (attempt 1): [matching checks and missed periods] expected 1 deliveries to com.lattice.journal::add-entry, got 0	aggregate	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	1 / 3	3 / 3	0 / 3
aggregate-21	On the first of each month at 8am, count last month’s Journal entries by mood and save the counts in Notes. Oracle states and failures Oracle states matching last calendar month by mood nonmatching current or earlier month boundary first and last March dates empty entries First failure per condition Luna · without table definitions (attempt 2): [matching last calendar month by mood] expected 1 deliveries to com.lattice.notes::add-note, got 0	aggregate	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	2 / 3
aggregate-22	Every Sunday at 4pm, email me the number of dated itinerary items for each Trips plan starting in the next 30 days. Oracle states and failures Oracle states matching trips with separate item counts nonmatching past or day thirty-one boundary day thirty with zero itinerary items empty trips First failure per condition Luna · without table definitions (attempt 1): refused a feasible request	aggregate	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	0 / 3
aggregate-23	Every Monday at 8am, total the settlement payments recorded in Shared Expenses over the past seven days by currency, and email me the separate totals without converting them. Oracle states and failures Oracle states matching payment-only currency totals nonmatching purchases and old payments boundary lower date permits rolling or full calendar dates empty expenses First failure per condition Luna · without table definitions (attempt 1): [matching payment-only currency totals] expected 1 deliveries to com.lattice.email::send-email, got 0	aggregate	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	0 / 3
aggregate-24	Every Sunday at 8am, email me a list of Pantry stock lots that still have some quantity left and reach their best-before date in the next seven days, grouped by product. Oracle states and failures Oracle states matching stock lots grouped by product nonmatching expired, late and undated lots boundary seventh day and fractional positive amount empty pantry First failure per condition Luna · without table definitions (attempt 1): refused a feasible request	aggregate	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	0 / 3
batch-01	Every evening at 9pm, gather all tasks that were completed in Todoist today and save them into a daily accomplishments note in Notes. Oracle states Oracle states live completions today, not old completed snapshots no new completions, despite completed stored tasks midnight maintenance discards previous evening buffer without a note creation and update are not completion late yesterday is not today All three conditions passed every attempt.	batch	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
batch-02	Every Sunday night at 8pm, count how many new feed items were added to Feed Reader this week and email me a digest of their headlines. Oracle states Oracle states new arrivals with old and recent RFC822 publication dates no arrivals despite stored articles one digest then cleared window buffering does not send mail before the scheduled flush guarded extra flushes must preserve arrivals All three conditions passed every attempt.	batch	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
batch-03	Collect new Feed Reader headlines during the day and email them to me in one digest at 6pm each weekday. Oracle states Oracle states today's new headlines; backlog and edits excluded no new items two weekdays: one digest each, no repeats All three conditions passed every attempt.	batch	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
batch-04	Gather all new GitHub Inbox threads received each day and send me one email at 5:30pm with their titles and links. Oracle states Oracle states today's new threads with titles and links quiet day two days: no repeats All three conditions passed every attempt.	batch	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
batch-05	Collect the Strava activities I record during the week and send me a single email every Sunday at 8pm listing their activity types, distances, and elapsed times. Oracle states and failures Oracle states week's activities: types, distances and elapsed (not moving) times no activities following week only First failure per condition Sol · full pipeline (attempt 2): [following week only] wrong distance or elapsed time for Walk	batch	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	1 / 3	3 / 3
batch-06	Collect Todoist tasks I delete during the day and email me one list of their titles every evening at 7pm. Oracle states Oracle states one list, only deletion events today separate days with actual scheduled flushes empty batch All three conditions passed every attempt.	batch	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
batch-07	Gather Grocy stock-consumption events on Tuesdays and Fridays, then at 8pm on each of those days email me one list of the products and amounts used. Oracle states and failures Oracle states Tuesday excludes Monday backlog Friday excludes Thursday and stocks empty eligible day First failure per condition Luna · without table definitions (attempt 1): [Tuesday excludes Monday backlog] wrong/missing labeled \brice\b: 82.5 expected 2.5	batch	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	2 / 3
batch-08	Save up Splitwise settlement payments as they’re added and email me one list of their descriptions and amounts at 6pm on the last day of each month. Oracle states Oracle states 31-day month, only settlements leap February last day non-leap February last day backdated payment, unobserved current-month backlog and update-only decoy February early flush must preserve the correct last-day stream populated 30-day month empty month not last day stays silent All three conditions passed every attempt.	batch	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
batch-09	Collect my Habits streak milestones during the week and email me one list of the habit names and streak lengths every Sunday at 7pm. Oracle states Oracle states matching buffered milestones nonmatching habit checks boundary no report before Sunday flush empty week boundary next week has no old carryover All three conditions passed every attempt.	batch	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
batch-10	Collect packing checklist items I add to Trips during the week and send me one email every Friday at 6pm listing the new items under each destination. Oracle states and failures Oracle states matching created_at additions with old backlog nonmatching checked old backlog boundary checked new item remains an addition empty packing lists First failure per condition Sol · full pipeline (attempt 1): [nonmatching checked old backlog] empty summary must say zero/none Luna · without table definitions (attempt 1): refused a feasible request	batch	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	2 / 3	0 / 3
enrich-01	When I receive an email in my Personal Email from a new sender, add them to my contacts. Oracle states Oracle states first message, then same sender again existing contact previous personal inbox sender work instance only All three conditions passed every attempt.	enrich	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
enrich-02	When the weather forecast shows rain, check my calendar for outdoor events today and send a warning to my Personal Email. Oracle states and failures Oracle states rain + park event today sunny rain, nothing planned today rain at other outdoor locations First failure per condition Luna · full pipeline (attempt 1): [rain + park event today] expected 1 deliveries to com.lattice.email#personal::send-email, got 0 Luna · without table definitions (attempt 1): [rain + park event today] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	enrich	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	1 / 3	3 / 3	0 / 3
enrich-03	When I save a new recipe in Recipe Box, check what ingredients we already have in Grocy and put any missing items onto my Todoist grocery list. Oracle states and failures Oracle states missing lentils and carrots, milk already stocked all ingredients stocked with sufficient quantities catalog entry without stock is missing only newly saved recipes ingredients omitted by event but stored in recipe First failure per condition Sol · full pipeline (attempt 2): [missing lentils and carrots, milk already stocked] need exactly one shopping item for carrots Luna · without table definitions (attempt 2): [ingredients omitted by event but stored in recipe] wrong number of grocery item tasks	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	1 / 3	2 / 3
enrich-04	When an email arrives from someone who is saved in my Contacts, save a copy of their message into Notes with their full name. Oracle states Oracle states known sender and full message unknown sender display-name sender syntax All three conditions passed every attempt.	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
enrich-05	When someone opens a new food package in Grocy, if its expiration date is within 5 days, capture a priority task in Todoist to finish it. Oracle states and failures Oracle states five-day boundary for opened lot six days away despite other expiring lots expires today already expired lot may count as within five days missing lot First failure per condition Luna · without table definitions (attempt 2): refused a feasible request	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	1 / 3
enrich-06	When a new GitHub Inbox thread arrives, look up its repository name in my Notes and add any matching setup instructions to a Todoist task for the thread. Oracle states Oracle states matching setup note no matching instructions: thread task or nothing updates and reads are not new threads All three conditions passed every attempt.	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
enrich-07	When a new email arrives from someone in Contacts, create a Todoist task titled with their name and the email subject. Oracle states Oracle states known sender unknown sender display-name sender syntax sent mail is not an arrival All three conditions passed every attempt.	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
enrich-08	When an email arrives from someone in Contacts, look up their phone number and create a Tasks item to call them back with that number in the title. Oracle states Oracle states known sender with phone another exact address display address sender known sender without phone sent is not received unknown sender All three conditions passed every attempt.	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
enrich-09	When a Calendar event is updated, look for a Contact whose name appears in its title and email me the event title and that contact’s email address. Oracle states Oracle states title name match different name and event no name match creation is not update All three conditions passed every attempt.	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
enrich-10	When I consume a Grocy product, find saved Recipe Box recipes that list it as an ingredient and email me up to three recipe titles. Oracle states Oracle states quantity-prefixed ingredient strings only one matching recipe at most three among five eligible recipes no recipes for consumed product stock purchase is not consumption All three conditions passed every attempt.	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
enrich-11	When I add a book to Goodreads, look for Notes containing its title and email me the matching note titles. Oracle states and failures Oracle states book title in note content book title in note title all matching notes, not only first empty related notes rating is not addition First failure per condition Luna · full pipeline (attempt 3): [book title in note content] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	enrich	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	2 / 3	3 / 3	3 / 3
enrich-12	When I check off an item on a Trips packing checklist, add a line to that day’s Journal entry saying what I packed and for which trip. Oracle states and failures Oracle states matching checked item with trip lookup nonmatching itinerary addition boundary quantity and distinct trip empty: no rows or triggering event First failure per condition Luna · without table definitions (attempt 1): [matching checked item with trip lookup] missing trip identity	enrich	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	0 / 3
enrich-13	When my Exercise habit reaches a 14-day streak, count the workouts I logged during those 14 days and append the count to today’s Journal entry. Oracle states and failures Oracle states matching fourteen days with lower-window decoy nonmatching other habit and week unit boundary zero workouts at valid milestone empty: no rows or triggering event First failure per condition Sol · full pipeline (attempt 2): [matching fourteen days with lower-window decoy] expected 1 deliveries to com.lattice.journal::add-entry, got 0 Luna · without table definitions (attempt 1): [matching fourteen days with lower-window decoy] wrong/missing /workout\|activit\|exercise/: expected 2, got 1	enrich	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	1 / 3	2 / 3
enrich-14	When I add a Journal entry tagged “reading,” look up books I finished in Reading Log that day and append their titles to the same entry. Oracle states and failures Oracle states matching books on exact entry date nonmatching reading-like tag boundary finished book now rereading empty matching dated book lookup empty: no rows or triggering event First failure per condition Luna · full pipeline (attempt 1): [boundary finished book now rereading] expected 1 deliveries to com.lattice.journal::add-entry, got 0 Sol · full pipeline (attempt 1): [boundary finished book now rereading] expected 1 deliveries to com.lattice.journal::add-entry, got 0 Luna · without table definitions (attempt 1): [boundary finished book now rereading] expected 1 deliveries to com.lattice.journal::add-entry, got 0	enrich	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	1 / 3	0 / 3	0 / 3
enrich-15	When I record a settlement payment in Shared Expenses, check whether its date falls during one of my Trips and, if it does, add the destination and payment amount to that day’s Journal entry. Oracle states Oracle states matching settlement in trip nonmatching purchase and outside dates boundary inclusive trip end empty trips for real payment empty: no rows or triggering event All three conditions passed every attempt.	enrich	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
fanout-01	Every morning at 8 AM, check my contacts for today's birthdays, create a calendar event for each, and send a greeting email from my Personal Email. Oracle states Oracle states two birthdays with tomorrow decoy no birthday today birthday without email still gets calendar event All three conditions passed every attempt.	fanout	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
fanout-02	When I add a contact tagged `client', create a task to send them an introductory email and add a follow-up reminder to my calendar for one week later. Oracle states Oracle states client marker in contact notes no client marker case-insensitive tag list without email All three conditions passed every attempt.	fanout	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
fanout-03	When I receive an email in my Work Email with `meeting' in the subject, create a calendar event and a preparation task. Oracle states Oracle states work meeting with explicit date subject and instance filters uppercase meeting without date All three conditions passed every attempt.	fanout	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
fanout-04	If I log a workout on Strava lasting more than 60 minutes, put it on my Calendar and create a Todoist task to do twenty minutes of stretching. Oracle states Oracle states 3601 seconds even with shorter moving time strict sixty-minute boundary ride also counts as a workout All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-05	When a new notification thread comes in on GitHub Inbox, show a desktop notification and put a high-priority to-do in Todoist so I don't lose track of it. Oracle states Oracle states new review thread updates and reads are not new threads a mention also creates an urgent task All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-06	Whenever I move a book to currently-reading on Goodreads, block out 30 minutes on my Calendar every weekday evening and send me a desktop reminder. Oracle states and failures Oracle states weekday evening reading block finishing or rating is not starting another newly started book First failure per condition Luna · full pipeline (attempt 2): refused a feasible request Luna · without table definitions (attempt 2): refused a feasible request	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	2 / 3	3 / 3	1 / 3
fanout-07	When a new calendar entry is scheduled on my Calendar, create a preparation task in Todoist due one day before and show a confirmation notification on my screen. Oracle states Oracle states dated calendar entry updated events do not create preparation tasks year-boundary date arithmetic recurrence start date when date omitted All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-08	Whenever a new recipe is saved, create a Todoist task with its title and a Calendar event for the same day titled “Try: ” followed by the recipe name. Oracle states Oracle states saved recipe edits and removals are not saves another recipe on another day All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-09	If I mark a Goodreads book as read, create a note with the title and finish date, then email me a copy of that note. Oracle states Oracle states finished book: note, then email copy other book events, and an unrelated note, are not emailed year-end finish date All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-10	When a Grocy stock lot is purchased, create a Tasks reminder with the product and quantity, and also send me a notification with the same details. Oracle states Oracle states purchased lot: product name by product_id and amount catalog, consumption and opening are not purchases fractional amount All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-11	Whenever a Splitwise expense is edited, email me the changed expense details and create a note titled “Updated expense” with the same information. Oracle states Oracle states edited expense additions and deletions are not edits large decimal cost All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-12	When I add a Contact with a birthday, save a note with their name and birthday and add a yearly Calendar event for it. Oracle states Oracle states birthday contact another month/day unknown birthday or existing contact update All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-13	When I clear a Goodreads rating, send me a notification with the book title and create a Todoist task to decide on a new rating. Oracle states Oracle states zero clears rating another cleared book positive rating or unrated addition All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-14	When I remove a Calendar event, save its title in Notes and create a Tasks item called “Check plans for” followed by the title. Oracle states Oracle states removed event title another removed title update/create are not removal All three conditions passed every attempt.	fanout	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
fanout-15	When my Read habit in Habits reaches a seven-day streak, append a celebration to today’s Journal entry and email me about the milestone. Oracle states Oracle states matching Read seven-day milestone nonmatching habit or milestone boundary seven weeks is not seven days empty: no rows or triggering event All three conditions passed every attempt.	fanout	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
fanout-16	When I create a trip in Trips, add an 8am Calendar event on its departure date titled “Leave for [destination]” and a Tasks item to finish packing due the day before. Oracle states Oracle states matching new trip nonmatching itinerary modification boundary one-day trip still has both actions empty: no rows or triggering event All three conditions passed every attempt.	fanout	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
fanout-17	When I write a Journal entry tagged “gratitude,” complete my Gratitude habit for that day and save a Notes item containing the entry date and mood. Oracle states Oracle states matching gratitude nonmatching substring tag boundary backdated journal checks that date empty: no rows or triggering event All three conditions passed every attempt.	fanout	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
fanout-18	When I log a ride in Workouts, check off my Cycling habit for that day and append the ride’s distance and time to that day’s Journal entry. Oracle states Oracle states matching ride nonmatching run and ride update boundary backdated recorded ride empty: no rows or triggering event All three conditions passed every attempt.	fanout	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
filter-01	When a task is marked as high priority, send a reminder to my Work Email. Oracle states Oracle states priority changed to high low or unspecified priority created high priority All three conditions passed every attempt.	filter	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
filter-02	When I add a bookmark tagged `read-later', create a task to read it. Oracle states Oracle states read-later among several tags substring and absent tag tag case and whitespace All three conditions passed every attempt.	filter	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
filter-03	When the weather forecast includes rain, send me a notification to bring an umbrella. Oracle states and failures Oracle states rain forecast clear or snow rain casing First failure per condition Luna · full pipeline (attempt 1): refused a feasible request Luna · without table definitions (attempt 1): [rain forecast] expected 1 deliveries to com.lattice.notifications::receive-notification, got 0	filter	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	0 / 3	3 / 3	0 / 3
filter-04	If anyone enters a household expense in Splitwise that comes out to over $100, send an email to my personal address right away. Oracle states Oracle states decimal USD group purchase strict threshold; JPY 9000 is under $100 EUR 900 converts to over $100 household scope is not schema-visible one cent above threshold without a group household settlement interpretation is open All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-05	Alert me with a pop-up on my screen only if I log a run on Strava that is longer than 15 km. Oracle states Oracle states 15001 meters 15km boundary and wrong sport fractional meter over boundary All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-06	Whenever I rate a book 5 stars in Goodreads, create a new note in Notes titled 'Favorites' with the book title and author. Oracle states and failures Oracle states five stars includes title and author zero is unrated, not five stars author omitted by rating event, available in stored edition First failure per condition Luna · without table definitions (attempt 1): [author omitted by rating event, available in stored edition] missing nia vale	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	1 / 3
filter-07	If any new story discovered by Feed Reader contains the phrase 'severe weather warning', pop up an immediate desktop alert. Oracle states and failures Oracle states phrase in synopsis exact phrase and new-item trigger case-insensitive phrase in title First failure per condition Luna · without table definitions (attempt 3): [phrase in synopsis] missing weather alert	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	2 / 3
filter-08	When I add a new contact with a phone number to Contacts, create a follow-up task in Tasks reminding me to send them an introductory message. Oracle states Oracle states new contact with phone phone required and edits filtered another new contact All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-09	If a new Feed Reader item has “rent increase” in its title or synopsis, email me the headline and original link. Oracle states Oracle states phrase in title; link is not the opaque guid phrase in synopsis, different case separated words, a substring inside another word, and edits synopsis-only match may be skipped or sent with its link All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-10	If a new Splitwise expense is in euros and costs more than €75, create a high-priority Todoist task to review it. Oracle states Oracle states EUR 120.00 (string comparison would miss it) 75.00 is not more than 75; other currencies; edits one cent over settlement inclusion is open All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-11	When a Grocy stock lot is consumed, notify me only if the amount used is at least 2 units. Oracle states Oracle states exactly 2 units used below 2, and other stock events fractional amount above 2 All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-12	If Goodreads reports that I cleared a book rating, create a note with the book title saying I removed its rating. Oracle states Oracle states rating cleared set ratings, and a new book that was never rated another cleared rating All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-13	If I add a book to my Goodreads to-read shelf with a rating of at least 4, email me its title. Oracle states Oracle states inclusive four stars five stars and shelf decoy below threshold or wrong event All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-14	If a GitHub Inbox thread changes its reason to a mention, save its title and link in Notes. Oracle states Oracle states reason mention on update opaque thread ID and second subject new mention or type is not changed mention reason All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-15	If I stock at least 6 units of a Grocy product in one new lot, email me the product and amount. Oracle states Oracle states six inclusive with distinct product/lot ids fractional above threshold below threshold and consumption are silent All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-16	If I remove a Calendar event with “dentist” in its title, create a Todoist task to book another appointment. Oracle states Oracle states dentist removal dentist within title other removal or creation All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-17	If I complete a Tasks item whose title starts with “Call,” make a note recording its title and completion date. Oracle states Oracle states call prefix completion timestamp differs from delivery day nonprefix or creation All three conditions passed every attempt.	filter	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
filter-18	When I write a Journal entry tagged “trip idea,” make a Tasks item called “Plan trip idea from [entry date]” due the following week. Oracle states Oracle states matching exact tag nonmatching substring tag boundary tag alongside another empty: no rows or triggering event All three conditions passed every attempt.	filter	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
filter-19	If I miss a scheduled Spanish Practice habit in Habits, create a medium-priority Tasks item to practice the next day. Oracle states Oracle states matching missed Spanish period nonmatching other habit boundary closed weekly partial period empty: no rows or triggering event All three conditions passed every attempt.	filter	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
filter-20	When I add lodging details to Trips, make a task to confirm check-in only if that trip starts within the next 14 days. Oracle states and failures Oracle states matching near-term lodging nonmatching far trip or activity boundary day fourteen empty: no rows or triggering event First failure per condition Luna · without table definitions (attempt 1): [matching near-term lodging] connection 0: no such column: trip_id	filter	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	0 / 3
filter-21	If I add a Journal entry tagged “anxious,” create a Tasks reminder for tomorrow to take a short break. Oracle states Oracle states matching anxious tag nonmatching mood alone or tag substring boundary other tags and backdated entry empty: no rows or triggering event All three conditions passed every attempt.	filter	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
filter-22	When a thread added to Code Inbox has “deployment” in its title, create a high-priority Tasks item to read it tomorrow. Oracle states Oracle states matching deployment title nonmatching body/repository and updates boundary embedded deployment text empty: no rows or triggering event All three conditions passed every attempt.	filter	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
forward-01	When I add a new contact, send them a welcome email from my Personal Email. Oracle states Oracle states contact with email contact without email contact edit is not an addition All three conditions passed every attempt.	forward	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
forward-02	When the weather forecast is updated, add today's forecast as a calendar event. Oracle states Oracle states updated forecast alert is not a forecast update zero Celsius forecast All three conditions passed every attempt.	forward	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
forward-03	When I complete a task, create a note logging the completion with a timestamp. Oracle states Oracle states completion timestamp creation is not completion missing optional timestamp All three conditions passed every attempt.	forward	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
forward-04	When I receive an email in my Work Email, send me a desktop notification with the sender and subject. Oracle states Oracle states work message other email instance two distinct work messages All three conditions passed every attempt.	forward	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
forward-05	When I add a calendar event, create a preparation task for it. Oracle states Oracle states new calendar event edit is not an addition title-only recurring event All three conditions passed every attempt.	forward	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
forward-06	When I save a new bookmark, create a note with the bookmark's title and URL. Oracle states Oracle states saved bookmark removal is not a save All three conditions passed every attempt.	forward	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
forward-07	Whenever I finish a book on Goodreads, please add a task to Todoist reminding me to write down a quick review. Oracle states Oracle states finished book with unrated zero started or rated is not finished a second finished book All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-08	Send a notification to my computer whenever a new recipe gets saved in Recipe Box. Oracle states Oracle states new recipe editing is not saving a new recipe another recipe All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-09	Whenever Feed Reader finds a new article on one of my subscribed blogs, create a task in Todoist with the article headline so I can read it later. Oracle states Oracle states headline becomes a Todoist task updates are not new discoveries headline-less entry may be skipped or given a placeholder old publication date still newly received All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-10	When I open an article link in Feed Reader, add a note in Notes containing the story title and link for future reference. Oracle states and failures Oracle states opened story discovering is not opening title/link resolved by local row ID First failure per condition Luna · full pipeline (attempt 2): [opened story] expected 1 deliveries to com.lattice.notes::add-note, got 0 Luna · without table definitions (attempt 1): [title/link resolved by local row ID] missing ocean report	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	2 / 3	3 / 3	0 / 3
forward-11	Whenever a brand new product is added to our pantry catalog in Grocy, send a notification to my desktop so I can review it. Oracle states Oracle states new pantry product new stock is not a brand new catalog product different product All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-12	Whenever Feed Reader finds edited content for an existing item, send me an email with the story title and publication date. Oracle states Oracle states edited item: title and RFC 822 publication date new and opened items are not edits another edit with a different date synopsis-only edit may be skipped or sent without a literal undefined All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-13	When a new stock lot is added in Grocy, create a note with the product name and the quantity that was stocked. Oracle states Oracle states stock unit amount and product name by product_id fractional stock amount catalog, consumption and opening are not new stock All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-14	Whenever I move a book to currently-reading in Goodreads, add a Todoist task to read it, using the book title as the task name. Oracle states Oracle states book moved to currently-reading adding, finishing and rating are not starting title with punctuation, author omitted by the event All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-15	When I record an activity in Strava, create a Calendar event titled with the activity type and its local start time. Oracle states Oracle states ride at local 12:15 although UTC is 20:15 early trail run edits and deletions are not new recordings All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-16	When I send an email, save a note with its subject and the names of the people I sent it to. Oracle states and failures Oracle states sent recipient display name address names and multiple recipients received mail is not sent First failure per condition Luna · without table definitions (attempt 3): [address names and multiple recipients] missing Ada Vale	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	2 / 3
forward-17	When I add a book to Goodreads, make a note with its title and the shelf I put it on. Oracle states Oracle states added to reading list current shelf not a custom shelf rating event is not an addition All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-18	Whenever I create a note, add a Todoist task called “Revisit” followed by the note’s title. Oracle states Oracle states new note punctuation and another title updates do not create revisits All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-19	When I create a task in Todoist, copy its title and due date into Tasks so it appears in both places. Oracle states Oracle states date-only deadline undated task stays undated updates are not new tasks All three conditions passed every attempt.	forward	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
forward-20	Whenever I log a walk in Workouts, check off my Walk habit in Habits for that day. Oracle states Oracle states matching walk nonmatching run and edit boundary backdated walk uses its date empty: no rows or triggering event All three conditions passed every attempt.	forward	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	3 / 3
forward-21	When I finish a book in Reading Log, append its title and my rating to that day’s Journal entry. Oracle states and failures Oracle states matching finished book matching sparse finish payload uses the stored rating nonmatching rating or start boundary backdated finish with unrated zero empty: no rows or triggering event First failure per condition Luna · without table definitions (attempt 1): [matching sparse finish payload uses the stored rating] wrong/missing book rating	forward	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	0 / 3
infeasible-01	When my phone battery drops below 15%, text my partner that I might be unreachable. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-02	Whenever a shared expense is added in Splitwise, look up that person's email address in my Contacts and send them a quick confirmation email. Oracle states and failures Oracle states empty workspace unrelated available event cannot supply the missing capability First failure per condition Sol · full pipeline (attempt 1): proposed a workflow for an infeasible request	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	0 / 3	3 / 3
infeasible-03	Whenever I mark a notification thread as read in GitHub Inbox, complete the corresponding task in Todoist. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-04	If an expense in Splitwise gets edited and the cost changes by more than $50, let me know with a desktop notification. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-05	Whenever an expense gets deleted from Splitwise, send an email to everyone involved and show an urgent pop-up notification on my screen. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-06	When a recipe in Recipe Box gets updated with a revised cooking time, update the scheduled dinner entry on my Calendar accordingly. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-07	Whenever a GitHub notification thread is about a review request, check if the person is in my Contacts and, if so, email me the thread link. Oracle states and failures Oracle states empty workspace unrelated available event cannot supply the missing capability First failure per condition Luna · full pipeline (attempt 2): proposed a workflow for an infeasible request Sol · full pipeline (attempt 1): proposed a workflow for an infeasible request	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	2 / 3	0 / 3	3 / 3
infeasible-08	Whenever I finish a run on Strava that's longer than 5 miles, send a text message to Mom letting her know I got back safe. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-09	If Grocy shows that we're completely out of milk or eggs, automatically reorder them on Instacart. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-10	Every time a new bill is posted in Splitwise where I owe money, pay my share immediately from my checking account using Venmo. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-11	Whenever I finish a book on Goodreads and give it 5 stars, post my review and rating to my Twitter profile. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-12	If an incoming email has 'Urgent' in the subject line, turn the living room smart lamp bright red. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-13	When I start cooking a dish in Recipe Box that takes over 45 minutes, play my Cooking Favorites playlist on Spotify. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-14	Whenever a marketing email arrives with an unsubscribe link, unsubscribe me automatically and permanently delete it from my inbox. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-15	Take a picture of my grocery receipts with my phone camera and automatically add all the purchased items into Grocy. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-16	When I save a recipe, post it to my family group chat so everyone can vote on dinner. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-17	If a new Feed Reader story mentions a company in my watchlist, buy $100 of its stock through my brokerage. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-18	When a new email arrives from a client, check their company’s website for their latest phone number and add it to Contacts. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-19	Whenever a contact's birthday is coming up, send them a birthday message on WhatsApp. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-20	Every night at 10pm, check my pantry and automatically add anything below its minimum stock level to an online supermarket cart. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-21	Every month, total what I spent across Splitwise, my credit card, and my bank account, then send me the combined amount. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-22	Every Sunday, tell me how many hours I spent reading my Goodreads books and which one I read the longest. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-23	Every Friday, compare my actual grocery spending from bank transactions with the pantry stock I bought in Grocy and email me any discrepancies. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-24	When a Splitwise expense is added, check my bank balance and tell me whether I can afford my share before emailing me. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-25	When I add a Calendar event with an address, check the live traffic and add enough travel time to the event. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-26	Whenever I rate a book on Goodreads, post the rating and my review to my public book blog. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-27	At 7am each day, print the day's Calendar events and Todoist tasks on my home printer. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-28	Save each new Todoist task to my Calendar with its due time and estimated duration. Oracle states and failures Oracle states empty workspace unrelated available event cannot supply the missing capability First failure per condition Sol · full pipeline (attempt 3): proposed a workflow for an infeasible request	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	2 / 3	3 / 3
infeasible-29	If an email arrives with an attachment, save the attachment into my Notes and label it with the sender's name. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-30	Whenever a task in Todoist is completed, add up the minutes I spent on it and include the total in a weekly report. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-31	If a recipe is edited, tell me which ingredients changed and whether I already have those changed ingredients in Grocy. Oracle states Oracle states empty workspace unrelated available event cannot supply the missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-32	If I miss a Calendar event, text my sister to tell her I’m running late. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-33	When I finish a ride on Strava, upload its route to Google Maps as a public list. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-34	Every Saturday at 9am, check tomorrow’s weather forecast and email me if I should move my planned run indoors. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-35	If a Splitwise expense includes the word “rent,” transfer my share from my bank account automatically. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-36	When a new Feed Reader podcast item appears, add it to my Spotify listening queue. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-37	Every Friday at 3pm, print my unfinished Todoist tasks on the office printer. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-38	When I save a recipe, upload a copy to my Google Drive dinner folder. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-39	When I finish a Todoist task, mark its matching item in Tasks complete too. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-40	When a Goodreads book gets a rating below 3, move it to my to-read shelf. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-41	If a Splitwise expense is deleted, restore it with its original participants and amounts. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-42	After I open a GitHub Inbox thread’s link, mark that thread read automatically. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-43	Whenever I purchase a new Grocy stock lot, subtract its cost from the matching Splitwise expense. Oracle states and failures Oracle states empty workspace unrelated event cannot supply missing capability First failure per condition Sol · full pipeline (attempt 3): proposed a workflow for an infeasible request	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	2 / 3	3 / 3
infeasible-44	When I remove a recipe from Recipe Box, also delete the Todoist task with the same title. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-45	Every Monday at 9am, archive last week’s sent emails so they leave my sent-mail view. Oracle states Oracle states empty workspace unrelated event cannot supply missing capability All three conditions passed every attempt.	infeasible	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-46	If a flight in Trips is delayed, check live airline availability and rebook me on the earliest alternative flight. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-47	When I write a sad Journal entry, start my comfort playlist on Spotify. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-48	When I reach a 30-day streak in Habits, text my brother the good news. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-49	After I log a ride in Workouts, upload its route to Komoot. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-50	On the day I return from a trip in Trips, set my home thermostat to 21°C at 5pm. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-51	Every Sunday, tell me exactly how many minutes I spent writing each Journal entry that week. Oracle states and failures Oracle states Refusal request; no seeded oracle states. First failure per condition Luna · full pipeline (attempt 2): proposed a workflow for an infeasible request	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	2 / 3	3 / 3	3 / 3
infeasible-52	When I attach a PDF ticket to a booking in Trips, extract my seat number and add it to the itinerary. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-53	Whenever I check off my Sleep habit, copy the number of hours I slept from Habits into that day’s Journal entry. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	3 / 3	3 / 3	3 / 3
infeasible-54	Every Friday, convert all my Shared Expenses settlements to dollars using the exchange rate saved with each payment and email me the total. Oracle states and failures Oracle states Refusal request; no seeded oracle states. First failure per condition Luna · full pipeline (attempt 1): proposed a workflow for an infeasible request Sol · full pipeline (attempt 1): proposed a workflow for an infeasible request	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	0 / 3	0 / 3	3 / 3
infeasible-55	Before each trip in Trips, total the weight of its packing checklist and warn me if it exceeds my airline’s baggage limit. Oracle states Oracle states Refusal request; no seeded oracle states. All three conditions passed every attempt.	infeasible	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Infeasible	3 / 3	3 / 3	3 / 3
instance-01	When I receive an email in my Work Email from `[address]' or containing the word `urgent', forward an alert to my Personal Email saying `Check your work inbox'. Oracle states Oracle states Alice matches without urgent neither condition or wrong instance Urgent in subject or body whole word urgent, not a substring All three conditions passed every attempt.	instance	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
instance-02	Every evening at 8 PM, count unread emails from both my Work Email and Personal Email, and if the combined total exceeds 20, send a summary breakdown notification. Oracle states and failures Oracle states twenty-one unread with per-instance breakdown exactly twenty is suppressed empty data all unread in personal First failure per condition Luna · without table definitions (attempt 1): [twenty-one unread with per-instance breakdown] expected 1 deliveries to com.lattice.notifications::receive-notification, got 0	instance	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
multi-source-01	Every Sunday, find contacts I haven't emailed from my Personal Email in 30 days and send me a reminder notification. Oracle states and failures Oracle states never emailed and stale vs recent and received-only all recently emailed from personal empty data just beyond thirty days First failure per condition Luna · full pipeline (attempt 3): [never emailed and stale vs recent and received-only] unexpected cara recent in com.lattice.notifications::receive-notification Luna · without table definitions (attempt 1): refused a feasible request	multi-source	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	2 / 3	3 / 3	0 / 3
multi-source-02	Every morning, cross-reference today's calendar events with my contacts to find meetings with people whose birthday is this week, and create a note reminding me to bring a card. Oracle states Oracle states meetings by names in title/description and calendar-week birthdays birthday next week or event tomorrow empty data All three conditions passed every attempt.	multi-source	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
multi-source-03	Every quarter, find contacts I haven't emailed from my Work Email in 90 days, check whether there is a pending task for each, and for those without one, create a follow-up task and email me the list. Oracle states and failures Oracle states stale contacts, prior tasks and other-instance decoys all recent or pending empty data ninety-day boundary First failure per condition Luna · without table definitions (attempt 1): refused a feasible request	multi-source	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
multi-source-04	Every Friday at 5 PM, generate a weekly review: count completed tasks, count newly saved bookmarks, and include today's weather—then send a formatted digest to my Personal Email. Oracle states and failures Oracle states weekly counts and latest weather today empty data only previous-week data and yesterday weather weather with empty task/bookmark aggregates First failure per condition Luna · without table definitions (attempt 1): [weekly counts and latest weather today] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	multi-source	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
multi-source-05	When a contact has a birthday coming up in the next 7 days, look up our past notes in Notes for gift ideas and create a Todoist to-do to buy a present. Oracle states Oracle states birthday window endpoints and person-specific gift notes birthday outside window or absent birthday still deserves a task without old gift ideas All three conditions passed every attempt.	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
multi-source-06	When a calendar event is created, check my Todoist tasks for ones due on the same date and email me the event title alongside those task titles. Oracle states Oracle states tasks due on the event date no tasks that day: event-only email or nothing updates and removals are not creations All three conditions passed every attempt.	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
multi-source-07	Whenever a recipe is saved, check Grocy for its listed ingredients and add a note containing the recipe name and which ingredients are already stocked. Oracle states and failures Oracle states stocked ingredients by product name ingredients re-read from the stored recipe nothing stocked: a none note or nothing edits and removals are not saves First failure per condition Luna · without table definitions (attempt 1): [stocked ingredients by product name] connection 0: no such table: stock	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	2 / 3
multi-source-08	When a GitHub Inbox thread arrives, look for the repository name in my Notes and send me a notification with the thread link and any matching note text. Oracle states Oracle states thread link plus matching note text no matching note: link only updates and reads are not arrivals All three conditions passed every attempt.	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
multi-source-09	When an email comes in from someone in Contacts, check Calendar for events with that person's name in the title during the next seven days and email me the sender and any matching event dates. Oracle states Oracle states matching event within seven days unknown sender another contact contact without upcoming events: sender-only email or nothing All three conditions passed every attempt.	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
multi-source-10	Every Sunday at 5pm, email me next week’s Calendar events that fall on days when I recorded a Strava activity, with the activity type beside each event. Oracle states Oracle states historical activity dates and weekdays no same-date match another match and multiple sport types next-week calendar endpoints All three conditions passed every attempt.	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
multi-source-11	Every Friday at 4pm, check Grocy for products expiring within two days and Recipe Box for recipes that use them, then make a note of the matching products and recipes. Oracle states and failures Oracle states interior and day-two expiry matches expired/today endpoints left open no matching recipes First failure per condition Sol · full pipeline (attempt 1): [no matching recipes] an empty report contains a nonzero count Luna · without table definitions (attempt 1): refused a feasible request	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	2 / 3	2 / 3
multi-source-12	Every weekday at 7am, email me one list of tasks due today from both Tasks and Todoist, labeling which app each task comes from. Oracle states and failures Oracle states both apps due today one app empty no tasks due today First failure per condition Luna · without table definitions (attempt 1): [both apps due today] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	1 / 3
multi-source-13	Every evening at 8pm, make a note listing today’s new Feed Reader headlines alongside today’s new GitHub Inbox thread titles, under separate headings. Oracle states and failures Oracle states arrival date differs from source dates today only, yesterday arrivals excluded empty daily inboxes First failure per condition Luna · full pipeline (attempt 3): [arrival date differs from source dates] missing Fresh parser	multi-source	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	2 / 3	3 / 3	3 / 3
multi-source-14	Every Friday at 8pm, save a Notes summary showing each day’s Journal mood alongside whether I checked off my Meditation habit that day. Oracle states and failures Oracle states matching date join, other habit decoy nonmatching checks without journal dates boundary journal entry without any check empty sources First failure per condition Sol · full pipeline (attempt 3): [nonmatching checks without journal dates] empty summary must say zero/none Luna · without table definitions (attempt 2): [matching date join, other habit decoy] expected 1 deliveries to com.lattice.notes::add-note, got 0	multi-source	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	2 / 3	1 / 3
multi-source-15	Every Sunday at 7pm, append a weekly Journal reflection listing my Workouts activity types and the titles of books I finished in Reading Log that week. Oracle states and failures Oracle states matching weekly exercise and finished books nonmatching old history boundary one source empty and rereading recorded finish empty sources First failure per condition Luna · full pipeline (attempt 2): [boundary one source empty and rereading recorded finish] missing Reread Star Sol · full pipeline (attempt 1): [boundary one source empty and rereading recorded finish] missing Reread Star Luna · without table definitions (attempt 1): [matching weekly exercise and finished books] expected 1 deliveries to com.lattice.journal::add-entry, got 0	multi-source	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	1 / 3	0 / 3	0 / 3
multi-source-16	Three days before each Trips departure, save a Notes outline with that trip’s dated itinerary and any Calendar events scheduled during its travel dates. Oracle states and failures Oracle states matching itinerary and overlapping calendar nonmatching departure day four boundary inclusive calendar travel endpoints empty trips First failure per condition Luna · without table definitions (attempt 2): [matching itinerary and overlapping calendar] missing flight Alpha	multi-source	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	1 / 3
schedule-01	Every morning at 8 AM, find contacts whose birthday is today and send each a greeting from my Personal Email. Oracle states Oracle states two birthdays today no birthdays today today without an email address All three conditions passed every attempt.	schedule	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
schedule-02	Every evening at 6 PM, find all overdue tasks and send a digest to my Personal Email. Oracle states and failures Oracle states overdue and completed/due-today decoys nothing overdue empty data First failure per condition Sol · full pipeline (attempt 2): [nothing overdue] unexpected Due today in com.lattice.email#personal::send-email Luna · without table definitions (attempt 3): [overdue and completed/due-today decoys] unexpected finished decoy in com.lattice.email#personal::send-email	schedule	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	2 / 3	2 / 3
schedule-03	Every morning at 7 AM, find today's calendar events and create a preparation task for each. Oracle states Oracle states two events today events on other days empty data All three conditions passed every attempt.	schedule	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	3 / 3
schedule-04	Every day at 9 AM, find bookmarks saved in the last 24 hours and email me a digest from my Personal Email. Oracle states and failures Oracle states rolling twenty-four hours only old bookmarks empty data First failure per condition Luna · without table definitions (attempt 1): [rolling twenty-four hours] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	schedule	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
schedule-05	Every first of the month, find notes tagged `todo' older than 30 days and send a cleanup reminder notification. Oracle states and failures Oracle states old todo notes with age and tag decoys no eligible note empty data First failure per condition Luna · without table definitions (attempt 1): refused a feasible request	schedule	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	1 / 3
schedule-06	Every morning, find notes created yesterday tagged `urgent' and create a task for each. Oracle states and failures Oracle states yesterday urgent with casing and date decoys no urgent notes yesterday empty data First failure per condition Luna · without table definitions (attempt 3): [yesterday urgent with casing and date decoys] expected 2 deliveries to com.lattice.tasks::add-task, got 0	schedule	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	2 / 3
schedule-07	Every weekday at 8 AM, check the weather forecast and if the temperature is below freezing, send me a notification. Oracle states and failures Oracle states latest below freezing zero is not below freezing empty data latest warm after cold First failure per condition Luna · without table definitions (attempt 1): [latest below freezing] expected 1 deliveries to com.lattice.notifications::receive-notification, got 0	schedule	9 installed apps Bookmarks Calendar Contacts Email (work) Email (personal) Notes Notifications Tasks Weather	Feasible	3 / 3	3 / 3	0 / 3
schedule-08	Every Friday at 5pm, check Grocy for any pantry items expiring in the next 3 days and add them to my Tasks list so we cook them this weekend. Oracle states and failures Oracle states today and three-day boundary, product-specific units nothing expiring soon empty pantry First failure per condition Luna · full pipeline (attempt 1): [today and three-day boundary, product-specific units] missing milk Sol · full pipeline (attempt 2): [today and three-day boundary, product-specific units] missing milk Luna · without table definitions (attempt 1): refused a feasible request	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	2 / 3	1 / 3	0 / 3
schedule-09	Every weekday at 8:30am, find all high-priority tasks in Tasks that are due today and send me a morning summary email. Oracle states and failures Oracle states core Tasks due today, not Todoist no qualifying tasks another high-priority task First failure per condition Luna · without table definitions (attempt 1): [core Tasks due today, not Todoist] unexpected completed decoy	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	1 / 3
schedule-10	Every weekday at 7:30am, check Goodreads for books on my currently-reading shelf and email me their titles. Oracle states Oracle states currently-reading shelf only nothing currently being read one book on Friday All three conditions passed every attempt.	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
schedule-11	On the 15th of each month at noon, check Grocy for products with a stock quantity of zero and create a Tasks item for each one. Oracle states and failures Oracle states products with no positive stock everything stocked empty catalog First failure per condition Luna · without table definitions (attempt 1): refused a feasible request	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	0 / 3
schedule-12	Every Wednesday at 6pm, find my unfinished Todoist tasks due in the next seven days and email me their titles and due dates. Oracle states Oracle states unfinished tasks due in the next seven days nothing due soon another week All three conditions passed every attempt.	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
schedule-13	On the first day of each month at 9am, find my GitHub Inbox threads that are review requests and create a Todoist task for each one. Oracle states and failures Oracle states review requests, not mentions or subject types no review requests two unread review requests First failure per condition Luna · without table definitions (attempt 1): [review requests, not mentions or subject types] expected 1–2 deliveries to org.fixture.todoist::add-task, got 0	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	0 / 3
schedule-14	Every morning at 7am, find Contacts with birthdays today and create a Tasks reminder for each one. Oracle states Oracle states birthday month/day across years leap birthday nobody today All three conditions passed every attempt.	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	3 / 3
schedule-15	Every Thursday at 5pm, find Recipe Box recipes whose cooking time is 25 minutes or less and email me their titles and cooking times. Oracle states and failures Oracle states inclusive duration boundary and seconds hours form is not minutes empty quick recipes First failure per condition Luna · full pipeline (attempt 2): [inclusive duration boundary and seconds] missing Quick rice Luna · without table definitions (attempt 2): [inclusive duration boundary and seconds] expected 1 deliveries to com.lattice.email#personal::send-email, got 0	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	1 / 3	3 / 3	2 / 3
schedule-16	At 6pm each day, find Grocy stock lots expiring tomorrow and make one note listing their products and quantities. Oracle states and failures Oracle states tomorrow lots, fractional stock units year-end tomorrow nothing tomorrow First failure per condition Luna · without table definitions (attempt 1): [tomorrow lots, fractional stock units] expected 1 deliveries to com.lattice.notes::add-note, got 0	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	0 / 3
schedule-17	Every Tuesday at noon, find Feed Reader items published the previous day and save a note with their titles and original links. Oracle states and failures Oracle states RFC 822 previous day publication timezone offset crosses midnight no yesterday items First failure per condition Luna · without table definitions (attempt 1): [RFC 822 previous day] expected 1–2 deliveries to com.lattice.notes::add-note, got 0	schedule	14 installed apps Reading Log Workouts Shared Expenses Todoist Feed Reader Recipes Pantry Code Inbox Notifications Email (personal) Calendar Tasks Notes Contacts	Feasible	3 / 3	3 / 3	0 / 3
schedule-18	Every night at 9:30, find my Habits scheduled for today that I haven’t checked off and append their names to today’s Journal entry under “Try tomorrow.” Oracle states and failures Oracle states matching due unchecked habits nonmatching all checked or off weekday boundary start today versus archive today empty habits First failure per condition Luna · without table definitions (attempt 1): [matching due unchecked habits] expected 1 deliveries to com.lattice.journal::add-entry, got 0	schedule	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	0 / 3
schedule-19	Every morning at 8, find Trips departing in exactly 10 days and create a Tasks reminder for each one to review its packing checklist, due three days before departure. Oracle states and failures Oracle states matching trip ten days ahead with day nine decoy nonmatching eleven days ahead boundary departure one-day trip empty trips First failure per condition Luna · without table definitions (attempt 1): refused a feasible request	schedule	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	1 / 3
schedule-20	Every Tuesday at 6pm, find the two shortest Recipes with cooking times of 20 minutes or less and make a Tasks item for each to consider for Wednesday dinner. Oracle states and failures Oracle states matching two shortest among eligible and decoys nonmatching too long and unknown boundary seconds and twenty minutes empty recipes fewer than two eligible recipes First failure per condition Luna · without table definitions (attempt 1): [matching two shortest among eligible and decoys] expected 2 deliveries to com.lattice.tasks::add-task, got 0	schedule	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	2 / 3
schedule-21	Every Wednesday at 7am, put a 7pm Calendar event titled “Habit: [name]” on my calendar for each Habits routine scheduled for that day. Oracle states and failures Oracle states matching Wednesday routine and offday decoy nonmatching archived or future routines boundary Wednesday-specific routine empty habits First failure per condition Luna · without table definitions (attempt 1): [matching Wednesday routine and offday decoy] expected 1 deliveries to com.lattice.calendar::add-event, got 0	schedule	18 installed apps Bookmarks Calendar Code Inbox Contacts Email Shared Expenses Feed Reader Habits Journal Notes Notifications Pantry Reading Log Recipes Tasks Trips Weather Workouts	Feasible	3 / 3	3 / 3	0 / 3

Requests are loaded from bench/tasks through the benchmark loader. Scores come from the Luna and Sol final3 runs’ results.rescored.jsonl files.

Benchmark

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

Oracle states

First failure per condition

Oracle states

Oracle states

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

Oracle states

Oracle states

Oracle states

Oracle states

First failure per condition

Oracle states

Oracle states

First failure per condition

Oracle states

Oracle states

Oracle states

First failure per condition

Oracle states

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

Oracle states

First failure per condition

Oracle states

Oracle states

Oracle states

Oracle states

Oracle states

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition

Oracle states

First failure per condition