Support Console Operations Solver
Use this skill when a task asks you to resolve support tickets, mobile support cases, mobile-data worklists, or enterprise export complaints using a shared support-console API.
Always produce only the JSON requested by the task's provided answer template.
Required Workflow
- Read the task prompt, every payload file, and the answer template.
- Read the task's environment access note for
<TASK_ENV_BASE_URL>and allowed endpoints. Use only read-only business endpoints; never call a judge endpoint. - Query the API for every target record. Do not infer IDs, customer records, lines, plans, bills, diagnostics, or enterprise data from naming patterns alone.
- Build one output object that exactly matches the answer template: same top-level keys, exact enum spellings, JSON booleans, strings for inapplicable IDs as
"", and numeric zeros as0.0when a numeric amount field is present. - Recompute every summary count from the decisions immediately before returning.
Useful lookup pattern:
BASE="<TASK_ENV_BASE_URL-without-trailing-slash>"
curl -sS "$BASE/api/catalog"
curl -sS "$BASE/api/search?q=<id-or-client-name>"
If a direct contact-center case endpoint listed in the access note is unavailable, use /api/search?q=<case_id> and take the result whose collection is cases. The catalog may also reveal equivalent public path names. Still use only public read-only records.
Ordering Rules
- Ticket batch decisions: preserve payload row order.
- Queue quality ticket decisions: preserve payload row order.
- Case decisions: sort by ascending
case_idunless the prompt explicitly says to preserve payload order. share_permissions: preserve the user order listed in the response requirements.- Output field order should follow the template for readability.
Ticket And Queue Quality Tasks
For each payload ticket:
- Fetch the ticket record and confirm the payload account/service values against API evidence.
- Find the account record, usually through
/api/search?q=<account_id>if a direct account endpoint is not listed. - Check for an active outage matching both
service_areaandservice_type. - For eligible non-outage tickets, fetch diagnostics and troubleshooting for the ticket.
Decision priority:
- Invalid account: if no account record exists for the payload account, set
FAILED, no diagnostic, no team, and useINVALID_ACCOUNTorINVALID_ACCOUNTblocker/route where that field exists. - Authentication failure: if account authentication shows failed login or failed recovery for the issue, set
FAILED, no diagnostic, no team, andAUTH_FAILED. - Suspended or ineligible account: set
FAILEDwith no diagnostic. Use the most specific blocker/route available from account status, line/account reason, and intake note. Overdue suspension routes toACCOUNTS_PAYABLEin queue-quality schemas that include route teams; generic ineligible-account schemas usually keep teamNONE. - Active outage: set
PENDING_ACTION, no diagnostic, no team, include the matchingoutage_idwhen the template has that field, and useOUTAGE_WAITorACTIVE_OUTAGEas applicable. - Resolved by troubleshooting: set
RESOLVED, teamNONE, blockerNONE, and routeAUTO_TROUBLESHOOTINGwhen present. - Not resolved by troubleshooting and root cause requires ownership: set
ESCALATEDand map the root cause to the appropriate escalation team.
Escalation mapping:
- Backbone, regional capacity, network saturation, or capacity root causes:
NETWORK_ENGINEERING; blockerNETWORK_CAPACITY. - Provisioning mismatch, stale provisioning, stale profile after move, or service profile mismatch:
TIER2_SUPPORT; blockerPROVISIONING_STALEwhen available. - Fiber drop damage, physical line faults, signal loss after line work, or premises/field repair causes:
FIELD_OPS; blockerPHYSICAL_LINE_FAULTwhen available. - Overdue billing suspension:
ACCOUNTS_PAYABLE; blockerOVERDUE_SUSPENSION. - Fraud suspension: use
FRAUD_SUSPENSION; route to a human/escalation team only if the template has an appropriate enum.
Diagnostic fields:
diagnostic_neededanddiagnostic_requiredare true only for eligible, non-outage tickets where diagnostics/troubleshooting are part of the resolution or escalation evidence.- Set latency issue true when latency is materially high or the issue/root cause is latency/packet-loss related. A practical threshold from the examples is around 100 ms, but account for the service type and notes.
- Set stability issue true when jitter is elevated, packet loss/signal loss is present, or the customer reports intermittent service.
- Set bandwidth issue true when measured bandwidth is materially below subscribed speed or the issue/root cause is slow speed/capacity. Compare diagnostics to
subscribed_mbps. - For invalid, auth-failed, suspended, or outage-wait cases, issue booleans are false.
Ticket summaries:
- Count
RESOLVED,PENDING_ACTION,ESCALATED, andFAILEDdirectly from decision statuses. - Count team keys such as
TIER2_SUPPORT,FIELD_OPS,NETWORK_ENGINEERING, andACCOUNTS_PAYABLEfrom the route/escalation team field, excludingNONE. tickets_requiring_customer_waitis the count of decisions waiting on an active outage.
Mobile Case And Data Recovery Tasks
For each case:
- Get the case record by case ID, using
/api/search?q=<case_id>if needed. - Fetch the customer, line, device, plan, and bill records referenced by the case and line.
- Choose the smallest operation that directly fixes the evidence-backed cause. Do not choose a billing or carrier update when a device setting alone explains the issue.
Action priority:
- Suspended line with overdue bill:
SEND_PAYMENT_REQUEST; secondaryRESUME_LINE_REBOOT; include the overdue bill ID and amount; final routeBILLING_RECOVERY. - Suspended line for non-billing reasons or contract ended with no self-service fix:
TRANSFER_HUMAN; final routeHUMAN_TRANSFER. - Airplane mode enabled:
TOGGLE_AIRPLANE_MODE. - Missing/unseated/inactive SIM with no service:
RESEAT_SIM. - Abroad/traveling with phone roaming disabled:
TOGGLE_ROAMING. - Abroad/traveling with phone roaming enabled but line roaming disabled:
ENABLE_LINE_ROAMING; carrier update required; final routeCARRIER_UPDATE. - Data usage at or above plan limit and the payload gives accepted refuel GB:
REFUEL_DATA; charge is accepted GB multiplied by the plan'sdata_refueling_price_per_gb; final routeDATA_RECOVERY. - Mobile data disabled:
TOGGLE_MOBILE_DATA. - Data saver enabled for slow data:
TOGGLE_DATA_SAVER. - Old or restrictive network mode such as 3G-only for slow data:
SET_NETWORK_MODE. - VPN connected for slow or blocked data:
DISCONNECT_VPN. - MMS/photo send failure with missing messaging permission:
GRANT_MESSAGING_PERMISSION; set permission tosms,storage, orsms_and_storagebased on missing flags. - MMS/data configuration missing, such as absent MMSC/APN while permissions are present:
RESET_APN_REBOOT. - Wi-Fi calling issue with Wi-Fi calling disabled:
TOGGLE_WIFI_CALLING. - If no evidence-backed self-service, billing, or carrier operation applies:
TRANSFER_HUMANfor unresolved active complaints; otherwiseNO_ACTION.
Mobile output fields:
secondary_actionisNO_ACTIONexcept for the overdue-bill recovery flow, where the follow-up is line resume and reboot.permissionisNONEexcept forGRANT_MESSAGING_PERMISSION.bill_idis blank except for billing recovery.charge_amount_usdis0.0except for overdue bill payment or data refuel.data_refuel_gbis the accepted refuel amount forREFUEL_DATA; otherwise0.0.carrier_update_requiredis true for carrier-side line changes such as enabling line roaming; otherwise false.
Route mapping:
- In contact-center schemas with
SELF_SERVICE, route device-only fixes and permission changes toSELF_SERVICE. - In data-recovery schemas, route
REFUEL_DATAtoDATA_RECOVERY, carrier-side line changes toCARRIER_UPDATE, device setting fixes toDEVICE_SETTING_FIX, and transfers toHUMAN_TRANSFER. - In billing schemas, route overdue recovery to
BILLING_RECOVERY.
Mobile summaries:
self_service_fixes,billing_recoveries,carrier_updates, andhuman_transferscountfinal_routevalues.data_refuel_casescountsREFUEL_DATA.device_setting_fixescounts data-worklist decisions whose route isDEVICE_SETTING_FIX.total_estimated_customer_charge_usdis the sum of allcharge_amount_usdvalues in the worklist.
Enterprise Export Complaint Tasks
For structured enterprise response packages:
- Parse the complaint for client name, product, severity, approximate incident ID, requested artifacts, and requested users.
- Query enterprise accounts, incidents, export runs, messages, and the SLA contract for the selected enterprise account.
- Select the incident by exact incident reference when present; otherwise match client, product, timing, and complaint summary.
- Filter export runs by selected incident and product/account. The failure window is the consecutive failed-run span relevant to the complaint: earliest failed date, latest failed date, and count of failed days.
backfill_daysnormally equals the failed-day count when messages or a subsequent successful run confirm manual backfill/recovery; otherwise mark the response as needing engineering review.- Derive
root_cause_categoryas a concise lowercase phrase from both failure codes and messages. Translate codes into plain language and enrich with message evidence; do not output raw codes unless the template explicitly asks for codes. - Set
contributing_alert_issuetoARCHIVED_ALERT_ROUTEwhen root-cause or alert evidence was posted to an archive/archived alert route;NONEwhen normal evidence shows no such issue;UNKNOWNwhen evidence is insufficient. - Use
severity,engineering_owner, andaccount_ownerfrom the selected incident or enterprise account records. - Determine SLA credit from the SLA contract and account-escalation messages. Apply the percent only when the failure window satisfies the trigger.
- Build response artifacts from the requirement naming style:
- Lowercase hyphen channel from the client/account name.
- Investigation folder using the client name and the failure month/year.
- Export failure report title using the client name and product/failure theme.
- For listed permission users, preserve the required order. Assign explicit permissions from requirements when provided. If only users are listed, use evidence-based roles: finance/account approvers get
view; engineering or artifact collaborators getedit; upload-only participants getupload_only. If roles are not discoverable, use the conservative pattern of first reviewer asviewand subsequent collaborator asedit. - Set response status:
NEEDS_FINANCE_REVIEWwhen an SLA credit must be handled.NEEDS_ENGINEERING_REVIEWwhen root cause, recovery, or backfill is incomplete.UNDER_INVESTIGATIONwhen the incident is still unresolved and required evidence is missing.READY_TO_SENDonly when all required evidence is complete and no finance or engineering review remains.
Final Validation
Before returning:
- Compare the output against the answer template key by key.
- Confirm every payload item has exactly one decision and no non-payload item appears.
- Check every enum value is one of the template's allowed values.
- Ensure inapplicable IDs are
"", notnull. - Recalculate all summaries from the final decisions.
- Return only valid JSON, with no markdown or explanation.