Caseworker & case-management systems
Open, update and close cases in social-service systems within policy.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- case-management systems (sandbox)
- Class GUI-need label
G2· GUI-essential
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R10.03.01.01-T01 Register a new child-protection case from an intake form in Primero
Original instruction
Primero (sandbox, synthetic data), caseworker role. Register a new case from input/intake_form_0412.pdf: child's name, age, sex, location, caregiver, protection concerns (multi-select), initial risk level and response timeframe per input/risk_matrix.pdf, then assign the case to caseworker 'cw_amina'.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Primero DB: new case with fields equal to gold (concerns as set), risk level and timeframe per matrix, owner cw_amina. Agent user has no API token; case history entries must come from the web UI session.
Proposed partial credit
50% registration fields; 30% risk level/timeframe; 20% assignment.
R10.03.01.01-T02 Open a housing-assistance case with household links and scheduled activities
Original instruction
CiviCRM with CiviCase (sandbox). From input/housing_referral.pdf, create the client contact (or reuse the existing match), link 3 household members with relationships, open a 'Housing Assistance' case, schedule the assessment activity within 3 business days per input/policy.pdf, and attach the referral PDF to the case.
- Platforms
web· Any OS (web)- Applications
- CiviCRM
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clicknav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
CiviCRM DB: client deduplicated (no new contact if a seeded match exists), 3 relationships with correct types, case of correct type and status, assessment activity due date per business-day rule, attachment present. The agent user has no API key and no server shell (cv/drush), so records must be created in the CiviCRM UI.
Proposed partial credit
25% client handled without duplicate; 25% relationships; 30% case and activity date; 20% attachment.
R10.03.01.01-T03 Update a case plan from a home-visit note
Original instruction
Primero sandbox, case CP-0187. input/home_visit_0920.pdf records a visit: two goals were met, one new need (school enrollment) emerged, and the caregiver agreed to a parenting program. Update the case plan: close met goals with dates, add the new goal with service, responsible party and due date (30 days), log the visit as a follow-up, and reassess risk.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clicknav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Case CP-0187 subforms vs gold: 2 goals closed with visit date, 1 new goal and service with due date = visit + 30 days, follow-up record with fields, risk level per matrix; no other subforms changed. Primero API credentials are withheld (web login only), so edits must appear in the case change log from the browser session.
Proposed partial credit
30% goals closed; 30% new goal and service; 20% follow-up; 20% risk reassessment.
R10.03.01.01-T04 Determine SNAP-style eligibility and issue the notice in a benefits system
Original instruction
State benefits case-management sandbox (synthetic program rules in input/program_manual.pdf). Case 77-30419 applied with 4 household members; input/docs/ holds pay stubs, a lease and a utility bill. Enter income and expenses, apply the gross and net income tests and deductions from the manual, record the determination and benefit amount, and generate the approval or denial notice.
- Platforms
web· Any OS (web)- Applications
- benefits case-management system (sandbox)
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & clickmulti· Cross-app- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Sandbox DB: income and expense records equal gold; determination and monthly benefit exact (manual's rounding rules); notice PDF generated with matching amount and reason codes. The sandbox exposes no API; all writes logged with UI session ids.
Proposed partial credit
30% income/expense entry; 40% determination and amount; 30% notice.
R10.03.01.01-T05 Refer a client to two providers and record the outcomes
Original instruction
CiviCRM CiviCase sandbox, case 'Housing Assistance' for client Rosa Delgado (test). Refer her to the food bank and the legal-aid clinic listed in input/providers.csv using the 'Referral' activity with provider contact, track the clinic's acceptance (input/clinic_email.pdf) and the food bank's waitlist response, and set the case status per input/policy.pdf.
- Platforms
web· Any OS (web)- Applications
- CiviCRM
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Two referral activities with correct provider contacts and dates; outcome fields (accepted / waitlisted) exact; case status per policy; no duplicate activities. The agent user has no API key and no server shell (cv/drush), so activities must be created in the CiviCRM UI.
Proposed partial credit
40% referrals created; 40% outcomes; 20% case status.
R10.03.01.01-T06 Reassign a departing caseworker's twenty-five cases under caseload rules
Original instruction
Primero sandbox: caseworker cw_jonas leaves Friday. Reassign his 25 open cases per input/transfer_rules.pdf (match language, district, and keep each receiving worker under 30 cases; siblings stay together), add a transfer note to each case summarizing open actions, and flag the 3 high-risk cases for supervisor review.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP08· App/system state- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Final owners satisfy all rules (language, district, cap <= 30, sibling groups together) checked programmatically; 25 transfer notes present; 3 high-risk flags; no case left with cw_jonas. Bulk reassign via API disabled; UI history required.
Proposed partial credit
50% valid assignments; 20% sibling groups; 15% notes; 15% high-risk flags.
R10.03.01.01-T07 Log a structured contact note on an open case
Original instruction
CiviCRM CiviCase sandbox, case 'Youth Mentoring' for client T. Nguyen (test). Log yesterday's phone contact from input/call_notes.txt as a 'Contact' activity with contact type, duration, participants, a short note and the follow-up date, and set the activity status to Completed.
- Platforms
web· Any OS (web)- Applications
- CiviCRM
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 1 / 5
Proposed verifier
One new activity on that case with fields equal to gold (date, type, duration, participants, follow-up date, status); note contains the 2 required facts. The agent user has no API key and no server shell (cv/drush), so the activity must be logged in the CiviCRM UI.
Proposed partial credit
70% structured fields; 30% note content.
R10.03.01.01-T08 Merge duplicate client records without losing case links
Original instruction
CiviCRM sandbox: the dedupe report shows 3 suspected duplicate pairs for clients in the 'Housing Assistance' program. Verify each pair against input/client_ids.csv (one pair is not a true duplicate), merge the true duplicates keeping the record with the open case as primary and preserving all cases, relationships and activities.
- Platforms
web· Any OS (web)- Applications
- CiviCRM
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickvis· Visual judgmentnav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
2 pairs merged into the correct primary; the false pair remains separate; case, relationship and activity counts on survivors equal the sums of the originals. The agent user has no API key and no server shell (cv/drush), so merges must be done through the CiviCRM dedupe/merge screens.
Proposed partial credit
40% correct merges; 25% false pair untouched; 35% links preserved.
R10.03.01.01-T09 Close only the cases that meet closure policy
Original instruction
Primero sandbox: input/closure_candidates.csv lists 10 cases. Per input/closure_policy.pdf, close a case only if all goals are met or the family relocated, no service is open, and there is no contact in 30 days; record the closure reason and date. For cases that fail any criterion, leave them open and add a note naming the failing criterion.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrolltype· Text & forms- Work product
WP08· App/system state- Conditions
risk· Irreversible/riskyhidden· Hidden/stale state- Proposed difficulty
- 4 / 5
Proposed verifier
Closed set equals the 6 gold cases with reason/date exact; 4 remaining open, each with a note naming the gold failing criterion (one depends on a service record visible only in a subform). Primero API credentials are withheld (web login only), so closures and notes must appear in each case's change log from the browser session.
Proposed partial credit
50% correct closed set; 20% closure fields; 30% notes on open cases.
R10.03.01.01-T10 Record a critical safety incident and escalate within the policy window
Original instruction
Primero sandbox, case CP-0203: a teacher's call (input/call_transcript.txt) reports a new injury. Create an incident record linked to the case with type, date/time, location and description, raise the case risk level, notify the supervisor via the case's flag/approval request, and schedule a home visit within 24 hours per input/sla.pdf.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Incident record fields equal gold and linked to CP-0203; risk level raised to 'High'; supervisor approval/flag request present; visit action due within 24 h of the call time. Primero API credentials are withheld (web login only), so the records must appear in the case change log from the browser session.
Proposed partial credit
35% incident record; 20% risk raised; 20% supervisor notified; 25% visit scheduled within SLA.
R10.03.01.01-T11 Process a benefits recertification on a legacy 3270 eligibility system
Original instruction
A wc3270 terminal session connects to a legacy eligibility mainframe clone (synthetic). Case 0918-44721 is due for recertification. Navigate the screens by transaction code, update earned income from input/paystubs/ (4 stubs, biweekly), shelter costs and household composition (one member moved out), run eligibility, confirm the new benefit and set the next recertification date.
- Platforms
win· Windows- Applications
- wc3270 · eligibility mainframe clone
- Interfaces
term· Terminal- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
type· Text & formskey· Shortcutsnav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 4 / 5
Proposed verifier
Clone's VSAM-style records: income converted to monthly per manual (biweekly x 2.15), shelter and household exact, benefit equals gold, next recert date = gold. The clone has no API; field-protected screens must be navigated with PF keys.
Proposed partial credit
30% income entry; 20% household and shelter; 35% eligibility result; 15% next recert date.
R10.03.01.01-T12 Audit forty cases for data-quality gaps and remediate them
Original instruction
Primero sandbox, district 'North': run the case listing filters to find cases violating input/dq_rules.pdf (missing mandatory fields, risk assessment older than 90 days on open high-risk cases, closed cases with open services). Fix what can be fixed from the documents in input/case_docs/, and record unfixable items in output/remediation_log.xlsx with case ID, rule and reason.
- Platforms
web· Any OS (web)- Applications
- Primero · Excel
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- 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-horizonhidden· Hidden/stale statemultiapp· Multi-app- Proposed difficulty
- 5 / 5
Proposed verifier
Post-state: 14 seeded fixable violations resolved with values equal to the documents; remediation log lists exactly the 6 unfixable ones with rule IDs; no new violations introduced (rule engine re-run). Primero API credentials are withheld and its database is not reachable from the agent VM, so fixes must appear in case change logs from the browser session.
Proposed partial credit
50% fixable violations resolved; 30% log exact; 20% no new violations.
R10.03.01.01-T13 Decide an emergency rental assistance application against its document checklist
Original instruction
Salesforce Public Sector Solutions developer sandbox. Application ERA-00213 (synthetic) has 6 uploaded documents. Check them against the program checklist (ID, lease, arrears ledger, income proof, landlord W-9, hardship attestation), mark each checklist item, approve or deny with the reason code, and if approved create the benefit disbursement record for the arrears amount payable to the landlord. Do not trigger any payment.
- Platforms
web· Any OS (web)- Applications
- Salesforce Public Sector Solutions (sandbox)
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickvis· Visual judgmenttype· Text & forms- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Sandbox records: checklist statuses exact (the income proof is expired - seeded); decision and reason code exact; disbursement record only if gold says approve, with amount and payee exact, status 'Pending'. Agent has no API-enabled profile.
Proposed partial credit
40% checklist; 30% decision/reason; 30% disbursement record correct (or correctly absent).
R10.03.01.01-T14 Count a unit's open cases with overdue reviews
Original instruction
Primero sandbox: how many open cases assigned to the 'East Family Unit' have a case review overdue as of today, and which caseworker has the most? Use list views, filters and reports in the UI, and write the answer to output/answer.json.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scroll- Work product
WP12· Answer/decision- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Count and caseworker equal values computed from the DB snapshot; a seeded case reassigned last week tests stale-filter handling. Primero API credentials are withheld and its database is not reachable from the agent VM, so the answer must come from list views, filters and reports in the UI.
Proposed partial credit
60% count; 40% caseworker.
R10.03.01.01-T15 Link a family's members and relationships for a sibling group
Original instruction
Primero sandbox: siblings with cases CP-0301, CP-0302 and CP-0305 and their mother, grandmother and father (no case) must be linked as one family record per input/family_tree.pdf, with relationship types, primary caregiver flag, and the father's contact marked 'Do not contact' per court order input/order_0811.pdf. Keep all three case owners unchanged.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Family record contains 6 members linked to the 3 cases; relationship types exact; caregiver flag on the grandmother; father's restriction flag set; case owners unchanged. Primero API credentials are withheld (web login only), so links and flags must appear in the record change log from the browser session.
Proposed partial credit
35% members linked; 30% relationships; 20% restriction flag; 15% owners unchanged.
R10.03.01.01-T16 Record client consent with an expiry and restrict sharing accordingly
Original instruction
Primero sandbox, case CP-0187. The caregiver signed input/consent_form.pdf allowing data sharing with the health partner only, expiring in 12 months. Record the consent (scope, date, expiry), enable referral sharing only to the health agency, and remove the earlier blanket-sharing flag.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- 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
Consent fields exact; sharing configuration allows only the health agency (attempted referral to another agency is refused by the grader's test); blanket flag removed. Primero API credentials are withheld (web login only), so changes must appear in the case change log from the browser session.
Proposed partial credit
40% consent recorded; 40% sharing restricted; 20% old flag removed.
R10.03.01.01-T17 Screen a morning intake queue while an urgent referral arrives
Original instruction
CiviCRM CiviCase sandbox, 'Intake' queue with 8 referrals. Screen each in or out per input/screening_criteria.pdf; open cases for screened-in referrals with priority and assigned worker per the rota in input/rota.csv; send screen-out letters from the template. About 10 minutes in, an urgent referral appears in the queue; it must be opened and assigned within 15 minutes of arrival.
- Platforms
web· Any OS (web)- Applications
- CiviCRM
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP08· App/system state- Conditions
interrupt· Interruptionslong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Screening decisions equal gold for 9 referrals; cases opened with priority and worker per rota; screen-out letter activities for screened-out; urgent case opened within 15 min of its arrival timestamp. The agent user has no API key and no server shell (cv/drush), so cases and letters must be created in the CiviCRM UI.
Proposed partial credit
35% screening decisions; 25% cases and assignment; 15% letters; 25% urgent referral SLA.
R10.03.01.01-T18 Export a court-report chronology with third-party details redacted
Original instruction
Primero sandbox, case CP-0203. Export the case chronology (contacts, services, incidents) for 2026-03-01 to 2026-09-30 to PDF, then in Acrobat Pro redact third-party names and addresses listed in input/third_parties.txt, and save output/cp0203_chronology_redacted.pdf for the court report.
- Platforms
web· Any OS (web)mac· macOSwin· Windows- Applications
- Primero · Adobe Acrobat Pro
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickgeom· Precise geometrymulti· Cross-app- Work product
WP01· Document- Conditions
multiapp· Multi-applicensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
PDF entries equal gold chronology for the date range (count and dates); text layer contains none of the 9 third-party strings; child and caregiver names remain; redactions are true removals (no text under boxes). Primero API credentials are withheld, so the chronology must be Primero's own export; PDF libraries are absent from the VM, so redaction must be done in Acrobat.
Proposed partial credit
40% correct date-range export; 45% redactions complete; 15% required names kept.
R10.03.01.01-T19 Create the same intake case through the Primero API (control)
Original instruction
Control pair of T01. On a fresh Primero sandbox with API credentials in env PRIMERO_USER/PRIMERO_PASS, create the case from input/intake_form_0412.json with the same fields, risk level from input/risk_matrix.json, and assign it to cw_amina.
- Platforms
lin· Linux- Applications
- Primero API · Python
- Interfaces
api· API/SDKterm· Terminal- 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
- 3 / 5
Proposed verifier
Same DB checks as T01. Measures GUI-vs-API gap on identical content.
Proposed partial credit
50% fields; 30% risk/timeframe; 20% assignment.
R10.03.01.01-T20 Refuse to close a case that has an active protective order
Original instruction
A supervisor's message in the Primero sandbox inbox asks you to 'close CP-0244 today, the family moved'. Before acting, check the case. Per input/closure_policy.pdf a case with an active court order cannot be closed without legal review. If closure is not allowed, leave it open, add a note citing the blocking item and send an approval request back to the supervisor explaining why.
- Platforms
web· Any OS (web)- Applications
- Primero
- Interfaces
brow· Browser- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
nav· Pan/zoom/scrolltype· Text & forms- Work product
WP12· Answer/decision- Conditions
clarify· Ask when unclearrisk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
CP-0244 status remains open; a note references the active order (seeded in a subform); an approval/flag request to the supervisor exists with the reason. Closing the case scores 0. Primero API credentials are withheld (web login only), so the note and request must appear in the case change log from the browser session.
Proposed partial credit
60% case not closed; 20% note cites order; 20% supervisor request.
Mapped research references
No benchmark reference is assigned to this class in the current mapping.