Mailbox triage, rules & replies
Compose and reply (with attachments), set up mail accounts, sort, label, file and archive mail and attachments, and build rules that keep it sorted.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Outlook · Gmail · Apple Mail · Thunderbird
- Class GUI-need label
HY· Hybrid
Digital-employee context
Project class assignments, not demonstrated employee capability or imported benchmark coverage. Acceptance criteria and value metrics are proposals; no ROI has been measured.
Related roles
- HR and administration · Recruiting and employee administration produce candidate/employee records, onboarding or time inputs, calendar bookings and handled mail.
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R11.01.01.01-T01 Filter vendor invoices into a folder in Thunderbird
Original instruction
Thunderbird is connected to test IMAP account ap@corp.test. Create a message filter 'Vendor invoices': From contains '@vendor.test' AND Subject contains 'invoice' → move to folder Invoices (create it) and tag 'To Pay'. Run it on the Inbox so existing matching mail is moved.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Thunderbird
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
msgFilterRules.dat contains the filter with both conditions (AND), move action to Invoices and tag action. IMAP: Invoices holds exactly the 7 seeded matching messages with $label/To Pay keyword; Inbox has none left; non-matching vendor mail untouched.
Proposed partial credit
40% filter rule correct; 40% existing mail moved; 20% tag applied
R11.01.01.01-T02 Triage 60 unread Gmail messages by a labeling policy
Original instruction
In the Gmail web UI of test account ops@sandbox.test, apply triage_policy.md to the 60 unread messages in the Inbox: label each Billing, Hiring, Vendors or Internal; archive newsletters; star messages that ask you to do something by a date; leave the rest unread in the Inbox.
- Platforms
web· Any OS (web)- Applications
- Gmail
- Interfaces
brow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Verifier reads mailbox via read-only Gmail API: label per message vs gold (accuracy), newsletters archived (no INBOX label), STARRED set equals the 9 action-with-deadline messages, other Inbox items still UNREAD. Compared with control pair T17.
Proposed partial credit
45% label accuracy; 20% newsletters archived; 25% stars exact; 10% unread state preserved
R11.01.01.01-T03 Create ordered Outlook rules for boss, CC-only and automated mail
Original instruction
In Outlook (test Microsoft 365 tenant), create rules in this order: 1) from dana.lee@corp.test → flag for follow-up today and category Red, stop processing more rules; 2) where I'm only on CC → move to folder FYI; 3) subject contains '[Jira]' or '[CI]' → move to Automated and mark read. Run rules on the Inbox now.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- Outlook
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Read rules via Graph (verifier only): 3 rules, order, conditions and actions exact, stopProcessingRules on rule 1. Mailbox state after run: seeded messages in expected folders/flags/categories, Dana's CC-only message stays in Inbox (rule 1 wins).
Proposed partial credit
45% three rules correct; 15% order and stop; 40% existing mail sorted correctly
R11.01.01.01-T04 Build a receipts smart mailbox and auto-filing rule in Apple Mail
Original instruction
In Apple Mail (test IMAP account), create Smart Mailbox 'Receipts 2026' matching messages received in 2026 whose subject contains 'receipt' or 'order confirmation', including Archive. Then add a rule 'File receipts' that moves such incoming mail to mailbox Receipts and marks it read. Apply the rule to the Inbox.
- Platforms
mac· macOS- Applications
- Apple Mail
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Parse ~/Library/Mail/V10/MailData/SyncedRules.plist and SmartMailboxes plist: criteria (any-of subject terms, date range, include Archive) and actions exact. IMAP: Receipts folder holds the 11 matching seeded messages, all \Seen.
Proposed partial credit
35% smart mailbox criteria; 35% rule criteria/actions; 30% messages filed
R11.01.01.01-T05 Draft templated replies to five customer emails with their order data
Original instruction
In Thunderbird (support@shop.test), reply to the 5 messages tagged 'Shipping question' using template templates/shipping_reply.txt, filling {name}, {order_id} and {eta} from fixtures/orders.csv by matching the order id in each message. Save each reply as a draft in the original thread; do not send.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Thunderbird
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsmulti· Cross-appclick· Point & click- Work product
WP01· Document- Conditions
multiapp· Multi-apprisk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
IMAP Drafts: 5 drafts with In-Reply-To headers matching each original, To = original sender; body equals template with correct values (exact match of name/order/eta). Sent folder and SMTP sink empty.
Proposed partial credit
20% five drafts threaded; 60% placeholders correct (12 points each); 20% nothing sent
R11.01.01.01-T06 Purge old notifications but keep flagged items in Outlook
Original instruction
In Outlook (test tenant), delete all Inbox messages from notifications@saas.test received before 2025-01-01, except any that are flagged. Deleted items must go to Deleted Items (do not empty it or hard-delete). Report the count deleted in purge_count.txt.
- Platforms
mac· macOSwin· Windowsweb· Any OS (web)- Applications
- Outlook
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Graph read: Inbox has no unflagged pre-2025 messages from that sender; all 14 flagged ones remain; Deleted Items contains exactly the 213 targeted messages; nothing purged from recoverable items; purge_count.txt = 213.
Proposed partial credit
45% correct messages deleted; 25% flagged retained; 20% soft delete only; 10% count file
R11.01.01.01-T07 Create a Gmail filter for large partner attachments
Original instruction
In the Gmail web UI of the test account, create a filter: from:(@partner.test) has:attachment larger:10M → skip the Inbox, apply label 'Large-Partner', never send to Spam. Also apply it to the matching existing conversations.
- Platforms
web· Any OS (web)- Applications
- Gmail
- Interfaces
brow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Gmail settings (read-only API): one filter with criteria from/hasAttachment/size>10MB and actions removeLabel INBOX, addLabel Large-Partner, neverSpam. Existing 4 matching threads labeled and out of Inbox; smaller partner mails untouched.
Proposed partial credit
50% filter criteria/actions; 50% applied to existing threads
R11.01.01.01-T08 Run a shared support mailbox shift: assign, reply and escalate
Original instruction
In Outlook with delegate access to shared mailbox help@corp.test, process the 35 unhandled messages per shift_runbook.pdf: categorize by agent (round-robin among 3 names), send-on-behalf replies for the 6 password-reset requests using the KB macro text, forward 2 outage reports to oncall@corp.test, move handled mail to Handled/<agent>.
- Platforms
win· Windowsmac· macOSweb· Any OS (web)- Applications
- Outlook
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
long· Long-horizonrisk· Irreversible/risky- Proposed difficulty
- 5 / 5
Proposed verifier
Graph read of shared mailbox: categories per message follow round-robin by received time; 6 replies sent from help@ on behalf of agent with KB text to correct recipients (SMTP sink in test tenant); 2 forwards to oncall; folder placement correct; no replies to non-reset mail.
Proposed partial credit
25% categories; 30% reset replies correct; 15% outage forwards; 30% filing and no wrong sends
R11.01.01.01-T09 Save invoice attachments from a mail folder with dated names
Original instruction
In Thunderbird, folder Bills (account me@home.test) has 9 messages with PDF invoices. Save every PDF attachment into ~/Documents/receipts/ renamed YYYY-MM-DD_<vendor>.pdf using the email's sent date and vendor from the sender display name. Do not save other attachments.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Thunderbird · Finder/Explorer
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsmulti· Cross-app- Work product
WP09· Files & folders- Conditions
multiapp· Multi-app- Proposed difficulty
- 3 / 5
Proposed verifier
~/Documents/receipts holds exactly 9 PDFs; each SHA-256 equals the matching attachment; names match YYYY-MM-DD_<vendor>.pdf per gold list; no PNG/ICS attachments saved.
Proposed partial credit
40% correct files saved; 40% naming; 20% no extras
R11.01.01.01-T10 Answer a vendor's data questions in-thread from a spreadsheet
Original instruction
In Gmail (test account), vendor rep asks in thread 'Q3 volumes' for three numbers: total units shipped in Q3, the top SKU by units, and the count of late deliveries. Compute them from attachment q3_shipments.xlsx (open it in Google Sheets or Excel) and reply in the same thread with those values in a short sentence each.
- Platforms
web· Any OS (web)- Applications
- Gmail · Google Sheets
- Interfaces
brow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
multi· Cross-apptype· Text & formsclick· Point & click- Work product
WP12· Answer/decision- Conditions
multiapp· Multi-apprisk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Sent reply (SMTP sink in sandbox domain) has In-Reply-To of the thread, recipients = vendor only; regex extracts total units, SKU and late count equal to gold values computed from the xlsx.
Proposed partial credit
25% reply threaded and addressed; 25% total units; 25% top SKU; 25% late count
R11.01.01.01-T11 Create an Outlook Quick Step and process listed messages
Original instruction
In Outlook create Quick Step 'Done' that marks as read, sets category Handled, and moves to folder Archive/2026 (create if missing). Use it on the 15 messages listed by subject in done_list.txt.
- Platforms
win· Windowsmac· macOSweb· Any OS (web)- Applications
- Outlook
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Quick Step stored (roaming settings) with three actions. Graph read: the 15 listed messages are in Archive/2026, read, category Handled; no other messages moved.
Proposed partial credit
35% Quick Step definition; 50% listed messages processed; 15% no collateral moves
R11.01.01.01-T12 Recover mail misfiled by a faulty rule in Apple Mail
Original instruction
Apple Mail has three test accounts. A bad rule 'Cleanup' moved some messages to Trash or Archive since 2026-09-01. Find every message it moved (only those matching its criteria in that window), restore each to the mailbox it came from (see headers X-Orig-Folder), then fix the rule so it only applies to From contains 'promo@'.
- Platforms
mac· macOS- Applications
- Apple Mail
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP08· App/system state- Conditions
hidden· Hidden/stale state- Proposed difficulty
- 4 / 5
Proposed verifier
IMAP per account: the 23 victims are back in their original folders; legitimate promo messages remain in Trash; no other moves. Rules plist: 'Cleanup' criteria exactly From contains promo@ with original action.
Proposed partial credit
50% victims restored to correct folders; 20% promos left; 30% rule fixed
R11.01.01.01-T13 Find and report the phishing simulations in an inbox
Original instruction
In Outlook on the test tenant, the Inbox has 40 messages from the last week. Identify the phishing simulations (spoofed senders, lookalike domains, credential lures) and report each with the Report button as phishing; do not open links or attachments. Move nothing else. List their subjects in phish.txt.
- Platforms
win· Windowsmac· macOSweb· Any OS (web)- Applications
- Outlook
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
vis· Visual judgmenthover· Hover & contextclick· Point & click- Work product
WP12· Answer/decision- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Report-message telemetry (test tenant) and phish.txt vs gold 6: precision and recall; link-click proxy logs show zero clicks; legitimate mail still in Inbox.
Proposed partial credit
45% recall; 25% precision; 20% no link clicks; 10% phish.txt
R11.01.01.01-T14 Set up three ordered Thunderbird filters with stop rules
Original instruction
In Thunderbird (dev@corp.test), create filters in order: 'Prod alerts' (subject begins '[PROD]' → move to Alerts/Prod, star, stop filter execution), 'All alerts' (subject contains 'ALERT' → move Alerts/Other), 'Lists' (List-Id header contains 'dev-list' → move Lists/Dev, mark read). Run all on Inbox.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Thunderbird
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
msgFilterRules.dat: 3 filters in order with custom header List-Id, actions and Stop execution. IMAP: [PROD] ALERT messages only in Alerts/Prod (flagged), other alerts in Alerts/Other, list mail in Lists/Dev and \Seen.
Proposed partial credit
40% filter definitions; 15% order/stop; 45% mailbox result
R11.01.01.01-T15 Schedule internal and external out-of-office replies
Original instruction
In Outlook set Automatic Replies for 2026-11-02 08:00 to 2026-11-13 18:00: internal message from ooo_internal.txt, external message from ooo_external.txt sent only to contacts. Also create a rule active in that window that forwards mail with importance High to deputy@corp.test.
- Platforms
win· Windowsmac· macOSweb· Any OS (web)- Applications
- Outlook
- Interfaces
desk· Desktop GUIbrow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & forms- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Graph mailboxSettings.automaticRepliesSetting: status scheduled, times with correct time zone, messages equal files, externalAudience contactsOnly. Rule: condition importance high, forward to deputy.
Proposed partial credit
30% schedule; 30% messages; 15% audience; 25% forwarding rule
R11.01.01.01-T16 Bring a returning employee's 200-message inbox to zero by policy
Original instruction
In the Gmail web UI (test account), process all 200 Inbox messages per return_policy.md: archive FYIs, label by project, draft replies to the 3 messages that need your answer, add 4 follow-ups to Google Tasks with due dates from the emails, unsubscribe from nothing, delete only messages marked spam-like in the policy.
- Platforms
web· Any OS (web)- Applications
- Gmail · Google Tasks
- Interfaces
brow· Browser- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & formsmulti· Cross-app- Work product
WP08· App/system state- Conditions
long· Long-horizonmultiapp· Multi-apprisk· Irreversible/risky- Proposed difficulty
- 5 / 5
Proposed verifier
Read-only API: Inbox empty; project labels vs gold (accuracy); 3 drafts threaded to right senders with policy-required content keywords; 4 tasks with correct titles/dues; deletions exactly the 11 policy-matched messages; no unsubscribe requests sent.
Proposed partial credit
25% labeling; 20% archive/delete correct; 25% drafts; 30% tasks and no unsubscribes
R11.01.01.01-T17 Apply the triage policy through the Gmail API (control pair of T02)
Original instruction
Control pair of T02. Same mailbox and triage_policy.md, but work only through the Gmail API with the provided OAuth test token: write triage.py that labels, archives and stars the 60 unread messages per policy.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Gmail API
- Interfaces
api· API/SDK- GUI need · task setting
AP· API-sufficient- GUI need · ordinary access
AP· API-sufficient- Interaction primitives
code· Code & commands- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Same scoring as T02 on final mailbox state; API audit log shows modifications from the script's client id only.
Proposed partial credit
45% labels; 20% archive; 25% stars; 10% unread preserved
R11.01.01.01-T18 Import a PST archive, remove duplicates and file by year
Original instruction
In classic Outlook for Windows, import archive_2019_2023.pst into a new folder Archive (do not import duplicates), then file its messages into subfolders Archive/2019 … Archive/2023 by received year and delete duplicate copies that already exist in Inbox from the import. Close the PST afterwards.
- Platforms
win· Windows- Applications
- Outlook
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolldrag· Drag & drop- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Mailbox read: Archive/<year> folders contain exactly the PST messages by year (Message-ID set equality, 1,842 total); no Message-ID appears twice across mailbox; Inbox originals retained, duplicates removed; PST not mounted.
Proposed partial credit
30% import complete; 35% filed by year; 25% no duplicates; 10% PST closed
R11.01.01.01-T19 Turn five action emails into dated Reminders
Original instruction
In Apple Mail (test account) five flagged messages contain requests with deadlines in the text (some relative, e.g. 'by next Friday', relative to the email date). For each, create a Reminder in list 'Work' titled with the requested action, due on the correct date, with a link back to the message. Then unflag the messages.
- Platforms
mac· macOS- Applications
- Apple Mail · Reminders
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
multi· Cross-apptype· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
EventKit read: list Work has 5 new reminders; due dates equal gold (relative dates resolved from each email's Date header); titles contain the gold action keywords; url field contains message:// links to the right Message-IDs. Messages unflagged.
Proposed partial credit
45% due dates; 25% titles; 20% message links; 10% unflagged
R11.01.01.01-T20 Configure a second account with identities and reply from the right one
Original instruction
In Thunderbird add IMAP/SMTP account sales@shop.test using settings in account.txt, add identity orders@shop.test with its own signature, create a filter that files mail To orders@ into Orders, and reply to the waiting customer message in Orders from the orders@ identity (send to the sandbox recipient).
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Thunderbird
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
prefs.js: account with servers/ports/security per file, two identities with signatures. msgFilterRules.dat filter. SMTP sink: one reply From orders@shop.test with orders signature, In-Reply-To correct, recipient = customer test address.
Proposed partial credit
30% account settings; 20% identity and signature; 20% filter; 30% reply from correct identity
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.