메뉴
Resolve SCN_003 CRM support-console tasks: ticket-batch resolution, mobile contact-center triage, and enterprise export-complaint response packages. Drive the remote console API, apply diagnostic floors + gating order, and emit the per-family JSON answer.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Reconcile MedBridge Sales Ops CRM tasks (B2B medical wholesale + implementation-services) against the remote Sales Ops API and emit the account-ready JSON answer. Covers QUOTE decision packages (EXW pricing, freight options, route risk, recommended mode, payment terms) and milestone RECONCILIATION packages (invoice/payment state, outstanding balance, revenue-recognition coverage, CRM follow-up routing, event/voucher linkage). Use whenever a task references MedBridge, quotes like Q-TR-*, RFQ-TR-*, opportunities like OPP-TR-*, milestones MS1/MS2/MS3, freight-quotes, revenue-journals, or the shared Sales Ops API base URL.
SCN_003 CRM support-operations solver. Resolve ticket batches (fixed-line diagnostics + gates), mobile contact-center triage, and enterprise export-complaint response packages against the shared support-console API. Apply exact metric thresholds, gate order, root-cause->team map, mobile action decision tree, and enterprise package derivation rules.
SCN_003 CRM support-console analyst. Resolve same-day ticket batches, mobile triage queues, mobile-data recovery worklists, and enterprise export-complaint response packages by querying the shared support-console API and applying fixed diagnostic thresholds, gating order, root-cause routing, and per-family decision trees.
Resolve CRM support-console tasks (SCN_003) — ticket-batch classification, mobile-data triage, and enterprise export-complaint response packages — against a remote support-console API. Use when a task asks to classify service tickets / queue snapshots, triage mobile data cases, or build an enterprise export-failure response package and references a support console API.
Solve ApexCloud Retention Operations analytics tasks against the remote Retention Operations API. Encode and reproduce the exact business rules, data-hygiene conventions, policy codes, ranking/precision rules, churn-model spec, and per-archetype output schemas. Use for any CRM retention task (renewal risk queue / save plan, QBR metrics packet, receivables + pipeline operations digest, churn model validation + outreach ranking, high-touch retention action board / watchlist).
ApexCloud Retention Operations — reusable rules and verified conventions for producing renewal-risk, QBR, receivables/pipeline, churn-model, and retention-board answers from the data API.
SOC 직업 분류 기준
| name | scn003-crm-support-console |
| description | Resolve SCN_003 CRM support-console tasks: ticket-batch resolution, mobile contact-center triage, and enterprise export-complaint response packages. Drive the remote console API, apply diagnostic floors + gating order, and emit the per-family JSON answer. |
Self-evolved skill. All thresholds marked (inferred) were derived by reasoning against the live API records (no gold answers). They are committed numeric estimates.
<remote-env-url> (ALWAYS use this; ignore any http://127.0.0.1:8057 in prompts).GET /health → {"ok": true}.GET /api/catalog (endpoints + record counts).| Endpoint | Returns |
|---|---|
GET /api/tickets | all tickets: ticket_id, account_id, service_type, service_area, subscribed_mbps, status, issue_summary |
GET /api/tickets/<ticket_id> | single ticket (carries subscribed_mbps + service_type) |
GET /api/accounts/<account_id> | status (Active/Suspended), authentication{last_login_status, account_recovery_status}, service_area, tier (404 {"error":"not_found"} = invalid account) |
GET /api/outages?service_area=<area> | list of {outage_id, active, eta_hours, service_types[], started_at} |
GET /api/diagnostics/<ticket_id> | PRE metrics: bandwidth_mbps, latency_ms, jitter_ms, root_causes[] |
GET /api/troubleshooting/<ticket_id> | POST metrics: post_bandwidth_mbps, post_latency_ms, post_jitter_ms, steps[] |
| Endpoint | Returns |
|---|---|
GET /api/cases / GET /api/cases/<case_id> | case_id, customer_id, line_id, device_id, issue_type, customer_location, summary |
GET /api/customers | customer_id, name, phone_number, status |
GET /api/lines/<line_id> | status (Active/Suspended), suspension_reason (OVERDUE_BILL / CONTRACT_ENDED / ""), roaming_enabled, data_used_gb, plan_id, device_id |
GET /api/devices/<device_id> | sim_status, signal_strength, airplane_mode, mobile_data_enabled, phone_roaming_enabled, data_saver_mode, network_mode_preference, vpn_connected, can_send_mms, mmsc_url_present, messaging_permissions{sms, storage}, wifi_calling_enabled, speed_test |
GET /api/plans/<plan_id> | data_limit_gb, data_refueling_price_per_gb, monthly_price_usd, name |
GET /api/bills | bill_id, customer_id, amount_due_usd, due_date, status (Paid/Overdue) |
| Endpoint | Returns |
|---|---|
GET /api/enterprise/accounts | enterprise_account_id, name, tier, account_owner, finance_owner |
GET /api/enterprise/incidents | incident_id, enterprise_account_id, product, severity, status, summary, engineering_owner, account_owner, received_at |
GET /api/enterprise/export-runs?incident_id=<id> | run_id, run_date, status (FAILED/SUCCEEDED), failure_code, exported_record_count |
GET /api/enterprise/messages?query=<text> | message_id, author, channel, created_at, body (substring search; use query=. to list all) |
GET /api/enterprise/sla/<enterprise_account_id> | monthly_export_credit_percent, credit_trigger, executive_contact |
Derived by flag-vs-metric reasoning across the 8 tickets that pass all gates (real root_causes, non-GENERATED_NOISE):
| Metric | Floor | Flag when | Cleared (post) when |
|---|---|---|---|
| bandwidth_mbps | 0.80 × subscribed_mbps (inferred) | bandwidth_mbps < 0.80·sub | post_bandwidth_mbps ≥ 0.80·sub |
| latency_ms | 100 ms (inferred) | latency_ms > 100 | post_latency_ms ≤ 100 |
| jitter_ms | 30 ms (inferred) | jitter_ms > 30 (= stability_issue) | post_jitter_ms ≤ 30 |
Evidence: latency 100 is the only round number in the viable window [82, 121) (resolved post-lat max = 82; escalated post-lat min with latency remaining = 121). Bandwidth 0.80 yields all-integer floors for every observed subscribed_mbps (300→240, 500→400, 750→600, 100→80, 200→160) and is consistent with both anchor tickets (TCK-5107 sub=300 pre 209<240 flagged, post 272≥240 cleared→RESOLVED; TCK-5184 sub=500 pre 318<400 flagged, post 332<400 remains→ESCALATED). Jitter 30 is the standard SLA threshold in [21,32) and flags the high-jitter resolved example (TCK-5107 pre 33.5) so auto-troubleshooting clears all three issue types.
Worked floor table (per subscribed_mbps):
| subscribed_mbps | bandwidth floor (0.80·sub) |
|---|---|
| 100 | 80 |
| 200 | 160 |
| 300 | 240 |
| 500 | 400 |
| 750 | 600 |
Evaluate each ticket in this strict order. The FIRST matching gate wins; gates 1–4 skip diagnostics.
GET /api/accounts/<account_id> → 404 {"error":"not_found"} (e.g. BAD-*).diagnostic_needed = false, outage_id = "".resolution_route = INVALID_ACCOUNT, escalation_team = NONE.key_blocker = INVALID_ACCOUNT, route_team = NONE.authentication.last_login_status == "FAILURE" OR account_recovery_status == "FAILURE".diagnostic_needed = false.resolution_route = AUTH_FAILED, escalation_team = NONE.key_blocker = AUTH_FAILED, route_team = NONE.account.status == "Suspended" (account exists, auth OK, but not Active → not eligible for diagnostics).diagnostic_needed = false.key_blocker:
OVERDUE_SUSPENSION (e.g. "overdue notice", "account hold", "suspended after overdue").FRAUD_SUSPENSION only if an explicit fraud/security indicator appears in the ticket text (none in train data; default suspended accounts are overdue).ACCOUNTS_PAYABLE (billing owns the suspension).
resolution_route = INELIGIBLE_ACCOUNT, escalation_team = ACCOUNTS_PAYABLE (billing owner; alternative if eval treats escalation_team as strictly for ESCALATED status → NONE).key_blocker = OVERDUE_SUSPENSION/FRAUD_SUSPENSION, route_team = ACCOUNTS_PAYABLE.PENDING_ACTION (customer pays) + ACCOUNTS_PAYABLE; chose FAILED per the hint's eligibility→FAILED rule. If status counts mismatch, reconsider PENDING_ACTION.GET /api/outages?service_area=<account.service_area>; if any outage has active == true AND ticket.service_type ∈ outage.service_types.diagnostic_needed = false, outage_id = <that outage_id>.resolution_route = OUTAGE_WAIT, escalation_team = NONE, issue flags (latency/stability/bandwidth) all false.key_blocker = ACTIVE_OUTAGE, route_team = NONE.tickets_requiring_customer_wait (= count of OUTAGE_WAIT tickets).diagnostic_needed = true. Fetch /api/diagnostics/<tid> and /api/troubleshooting/<tid>.bandwidth_issue = bandwidth_mbps < 0.80·subscribed_mbpslatency_issue = latency_ms > 100stability_issue = jitter_ms > 30post_bandwidth_mbps ≥ 0.80·subpost_latency_ms ≤ 100post_jitter_ms ≤ 30resolution_route = AUTO_TROUBLESHOOTING, escalation_team = NONE. train_004: key_blocker = NONE, route_team = NONE.resolution_route = ESCALATION, escalation_team = <root-cause team>. train_004: key_blocker = <root-cause blocker>, route_team = <root-cause team>.GENERATED_NOISE as the only root_cause = sentinel that the ticket was gated out (gates 1–4); never diagnose it. Real root_causes seen: CONFIGURATION_DRIFT, FIBER_DROP_DAMAGE, SIGNAL_LOSS, VOICE_PROFILE_STALE, BACKBONE_CAPACITY, PROVISIONING_STALE.| diagnostic root_cause | escalation_team | train_004 key_blocker |
|---|---|---|
FIBER_DROP_DAMAGE, SIGNAL_LOSS (fiber/signal) | FIELD_OPS | PHYSICAL_LINE_FAULT |
BACKBONE_CAPACITY (backbone/capacity) | NETWORK_ENGINEERING | NETWORK_CAPACITY |
PROVISIONING_STALE (provisioning) | TIER2_SUPPORT | PROVISIONING_STALE |
any BILLING_* (billing root cause) | ACCOUNTS_PAYABLE | (billing) |
CONFIGURATION_DRIFT, VOICE_PROFILE_STALE | typically auto-RESOLVED; if not cleared → TIER2_SUPPORT (fallback) | — |
RESOLVED / PENDING_ACTION / ESCALATED / FAILED: counts of each final_resolution_status.tickets_requiring_customer_wait: count of OUTAGE_WAIT (PENDING_ACTION via active outage) tickets.For each case, fetch case → customer → line → device → plan, and the customer's bills. Decide in this priority order (first match wins). customer_location "abroad" = traveler.
line.status == "Suspended")suspension_reason == "OVERDUE_BILL" AND a bill status == "Overdue":
final_route = BILLING_RECOVERY (train_002) — bill_id = the overdue bill_id, charge_amount_usd = bill.amount_due_usd. (train_005 has no billing action in its enum → if a train_005 case is an overdue suspension, use TRANSFER_HUMAN.)suspension_reason == "CONTRACT_ENDED" (or any non-bill):
final_route = HUMAN_TRANSFER (renewal/retention needed; cannot self-serve).issue_type == NO_SERVICE (not suspended): inspect device.sim_statussim_status == "missing" (e.g. after a commute) → RESEAT_SIM (train_002 only; not in train_005 enum → TRANSFER_HUMAN there).sim_status == "locked_pin" (SIM PUK-locked) → TRANSFER_HUMAN (HUMAN_TRANSFER).device.airplane_mode == true → TOGGLE_AIRPLANE_MODE (SELF_SERVICE).issue_type == MOBILE_DATA (data not working)line.data_used_gb > plan.data_limit_gb → REFUEL_DATA.
data_refuel_gb = customer_preferences[case_id].accepted_refuel_gb if the worklist provides it; else default = ceil(data_used_gb - data_limit_gb) (min 1.0) (inferred).charge_amount_usd = data_refuel_gb × plan.data_refueling_price_per_gb (2-decimal).final_route = DATA_RECOVERY (train_005); train_002 route = SELF_SERVICE.line.roaming_enabled == false (line-side roaming off; phone may be on) → ENABLE_LINE_ROAMING, carrier_update_required = true, final_route = CARRIER_UPDATE. (carrier/line-side update)phone_roaming_enabled == false (line roaming OK, phone toggle off) → TOGGLE_ROAMING (device self-service). train_002 final_route = SELF_SERVICE; train_005 final_route = DEVICE_SETTING_FIX. carrier_update_required = false.device.mobile_data_enabled == false (e.g. after a settings change) → TOGGLE_MOBILE_DATA. train_005 final_route = DEVICE_SETTING_FIX; train_002 SELF_SERVICE.device.airplane_mode == true → TOGGLE_AIRPLANE_MODE.issue_type == SLOW_DATAdevice.data_saver_mode == true → TOGGLE_DATA_SAVER (DEVICE_SETTING_FIX / SELF_SERVICE).device.network_mode_preference is an older mode (3g_only, 2g_only, etc.) → SET_NETWORK_MODE (upgrade to 4g_5g_preferred) (DEVICE_SETTING_FIX / SELF_SERVICE).device.vpn_connected == true → DISCONNECT_VPN (SELF_SERVICE / DEVICE_SETTING_FIX).issue_type == MMSdevice.mmsc_url_present == false (APN profile broken, e.g. after APN edit) → RESET_APN_REBOOT (SELF_SERVICE). (train_002 only.)device.messaging_permissions.storage == false and/or sms == false → GRANT_MESSAGING_PERMISSION (train_002 only).
permission field = the missing permission(s): storage if only storage missing; sms if only sms missing; sms_and_storage if both missing; NONE otherwise.can_send_mms == false but mmsc present + perms OK → consider TOGGLE_WIFI_CALLING or TRANSFER_HUMAN.)final_route = HUMAN_TRANSFER.| Action | train_002 final_route | train_005 final_route |
|---|---|---|
| RESEAT_SIM, TOGGLE_ROAMING, GRANT_MESSAGING_PERMISSION, DISCONNECT_VPN, RESET_APN_REBOOT, TOGGLE_AIRPLANE_MODE, TOGGLE_DATA_SAVER, SET_NETWORK_MODE, TOGGLE_MOBILE_DATA | SELF_SERVICE | DEVICE_SETTING_FIX (or SELF_SERVICE N/A in train_005) |
| REFUEL_DATA | SELF_SERVICE | DATA_RECOVERY |
| ENABLE_LINE_ROAMING | CARRIER_UPDATE | CARRIER_UPDATE |
| SEND_PAYMENT_REQUEST (+RESUME_LINE_REBOOT) | BILLING_RECOVERY | (not in enum → TRANSFER_HUMAN) |
| TRANSFER_HUMAN | HUMAN_TRANSFER | HUMAN_TRANSFER |
charge_amount_usd: REFUEL_DATA → refuel_gb × plan.data_refueling_price_per_gb; SEND_PAYMENT_REQUEST → bill.amount_due_usd; all other actions → 0.00 (2-decimal).carrier_update_required (train_005): true ONLY for ENABLE_LINE_ROAMING; false otherwise.data_refuel_gb (train_005): the refuel amount (1-decimal) for REFUEL_DATA; 0.0 otherwise.permission (train_002): storage/sms/sms_and_storage for GRANT_MESSAGING_PERMISSION; NONE otherwise.bill_id (train_002): the overdue bill_id for BILLING_RECOVERY; "" otherwise.secondary_action: RESUME_LINE_REBOOT for BILLING_RECOVERY; NO_ACTION for all other single-step actions.Inputs: complaint email (client name, product, incident ref) + response_requirements.json (required fields, permission_users_to_include, naming_style).
GET /api/enterprise/incidents (match incident_id + client).incident.enterprise_account_id (cross-check GET /api/enterprise/accounts by name == client).GET /api/enterprise/export-runs?incident_id=<id>.
start_date = run_date of the FIRST FAILED run.end_date = run_date of the LAST FAILED run.failed_days = count of consecutive FAILED runs. (Asteri: 2026-05-12 → 2026-05-14, 3. Quanta: 2026-05-25 → 2026-05-28, 4.)failed_days (the failed runs are backfilled; confirmed by the succeeding recovery run + messages saying "N days require manual backfill"). (Asteri 3, Quanta 4.)failure_code, lowercased, corroborated by message evidence. (Asteri: stale_credential; Quanta: staging_storage_quota.) Concise category string.ARCHIVED_ALERT_ROUTE | NONE | UNKNOWN): search GET /api/enterprise/messages?query=<client/incident>; if an alert message's channel contains "archive" (e.g. export-alerts-archive) → ARCHIVED_ALERT_ROUTE; if there is a clear alert in an active channel (e.g. data-platform) → NONE; if no message evidence → UNKNOWN. (Asteri: ARCHIVED_ALERT_ROUTE; Quanta: NONE.)GET /api/enterprise/sla/<enterprise_account_id> → monthly_export_credit_percent, applied when the credit_trigger condition is met (Asteri: "3 consecutive failed" → 15; Quanta: "critical export outage longer than 72 hours", 4 days=96h → 20). Integer percent.incident.severity (Critical/High/Medium/Low).incident.engineering_owner.incident.account_owner.-archive): Asteri → export-alerts. Otherwise use the channel from the engineering root-cause message (Quanta → data-platform). (inferred)<client-slug>-<failure_window.start_date>-investigation. Client slug = first two words of the account name, lowercased, hyphenated (drop Inc./Group): Asteri Retail Inc. → asteri-retail. → asteri-retail-2026-05-12-investigation. (inferred)<Client Name> Export Failure Report (Title Case, full client name minus Inc./Group): Asteri Retail Export Failure Report. (inferred)response_requirements.permission_users_to_include, in that exact order. Permission (view | edit | upload_only) (inferred): finance_owner (if in list, e.g. laura.brown) → edit (approves/adjusts SLA credit); other included users → view. (Asteri: [{user: laura.brown, permission: edit}, {user: jun.chen, permission: view}]. Alternative: laura.brown=view, jun.chen=upload_only.)READY_TO_SEND | NEEDS_FINANCE_REVIEW | NEEDS_ENGINEERING_REVIEW | UNDER_INVESTIGATION) (inferred):
UNDER_INVESTIGATION;NEEDS_ENGINEERING_REVIEW;NEEDS_FINANCE_REVIEW (finance must approve the credit before sending);READY_TO_SEND.Let D = set of diagnosed tickets (passed all gates; real root_cause). For t∈D: sub = subscribed_mbps; f_bw = 0.80·sub; f_lat = 100; f_jit = 30.
max(0, f_bw − pre_bandwidth_mbps) (total Mbps below the bandwidth floor before troubleshooting; only positive gaps). (inferred)total_tickets − |D|. (inferred)(post_bw < f_bw) + (post_lat > f_lat) + (post_jit > f_jit) summed over D. (inferred)max(0, f_bw − post_bandwidth_mbps).max(0, post_latency_ms − f_lat).max(0, post_jitter_ms − f_jit).{ticket_id, subscribed_mbps, bandwidth_floor=f_bw, pre_bandwidth_mbps, pre_bandwidth_shortfall=max(0,f_bw−pre_bw), pre_latency_ms, pre_latency_excess=max(0,pre_lat−f_lat), pre_jitter_ms, pre_jitter_excess=max(0,pre_jit−f_jit), post_* (same), pre_flags{bandwidth,latency,stability}, post_remaining_flags, root_causes, resolved_bool}. (inferred)root_cause → escalation_team (fiber/signal→FIELD_OPS, backbone/capacity→NETWORK_ENGINEERING, provisioning→TIER2_SUPPORT, billing→ACCOUNTS_PAYABLE). (inferred)ticket_id, account_id, final_resolution_status (RESOLVED|PENDING_ACTION|ESCALATED|FAILED), diagnostic_needed (bool), latency_issue (bool), stability_issue (bool, =jitter), bandwidth_issue (bool), outage_id (string; "" if none), escalation_team (NONE|TIER2_SUPPORT|FIELD_OPS|NETWORK_ENGINEERING|ACCOUNTS_PAYABLE), resolution_route (AUTO_TROUBLESHOOTING|OUTAGE_WAIT|ESCALATION|INELIGIBLE_ACCOUNT|AUTH_FAILED|INVALID_ACCOUNT).
batch_summary: RESOLVED, PENDING_ACTION, ESCALATED, FAILED, tickets_requiring_customer_wait.
ticket_id, final_resolution_status (RESOLVED|PENDING_ACTION|ESCALATED|FAILED), route_team (NONE|TIER2_SUPPORT|FIELD_OPS|NETWORK_ENGINEERING|ACCOUNTS_PAYABLE), key_blocker (NONE|ACTIVE_OUTAGE|INVALID_ACCOUNT|AUTH_FAILED|OVERDUE_SUSPENSION|FRAUD_SUSPENSION|NETWORK_CAPACITY|PROVISIONING_STALE|PHYSICAL_LINE_FAULT), diagnostic_required (bool).
queue_summary: FAILED, PENDING_ACTION, RESOLVED, ESCALATED, TIER2_SUPPORT, FIELD_OPS, NETWORK_ENGINEERING, ACCOUNTS_PAYABLE.
case_id, customer_id, line_id, primary_action, secondary_action, permission (NONE|sms|storage|sms_and_storage), bill_id ("" if n/a), charge_amount_usd (2-decimal), final_route (SELF_SERVICE|BILLING_RECOVERY|CARRIER_UPDATE|HUMAN_TRANSFER).
queue_summary: self_service_fixes, billing_recoveries, carrier_updates, human_transfers.
case_id, primary_action, secondary_action, data_refuel_gb (1-decimal; 0.0 n/a), charge_amount_usd (2-decimal), carrier_update_required (bool), final_route (DATA_RECOVERY|CARRIER_UPDATE|DEVICE_SETTING_FIX|HUMAN_TRANSFER).
worklist_summary: data_refuel_cases, carrier_updates, device_setting_fixes, human_transfers, total_estimated_customer_charge_usd.
incident_id, enterprise_account_id, root_cause_category, contributing_alert_issue (ARCHIVED_ALERT_ROUTE|NONE|UNKNOWN), failure_window{start_date,end_date,failed_days}, backfill_days, sla_credit_percent (int), severity (Critical|High|Medium|Low), engineering_owner, account_owner, channel_name, evidence_folder, report_title, share_permissions[{user,permission(view|edit|upload_only)}], response_status (READY_TO_SEND|NEEDS_FINANCE_REVIEW|NEEDS_ENGINEERING_REVIEW|UNDER_INVESTIGATION).
<remote-env-url>, never 127.0.0.1:8057.subscribed_mbps. Compute 0.80·subscribed_mbps per ticket.jitter_ms > 30.active==true AND service_type ∈ outage.service_types — a non-matching outage does not block.line.roaming_enabled (line-side → ENABLE_LINE_ROAMING, carrier update) vs device.phone_roaming_enabled (device toggle → TOGGLE_ROAMING, self-service). They are different fixes.OVERDUE_BILL → billing recovery (SEND_PAYMENT_REQUEST+RESUME_LINE_REBOOT); CONTRACT_ENDED/other → TRANSFER_HUMAN. Do not send a payment request for a non-bill suspension.mmsc_url_present==false → RESET_APN_REBOOT; storage/sms permission false → GRANT_MESSAGING_PERMISSION. Don't confuse the two.failure_window.end_date is the LAST FAILED run_date, not the recovery date. backfill_days = failed_days. sla_credit_percent comes from the SLA contract (per-account), applied when the trigger condition is met — it is NOT always 15 (Quanta=20).