Self Attempt 01

Support Console Skill — Task Group 003

Environment

• Base URL: http://34.46.77.124:8003 (remote only; never use localhost)
• Service: task_group_003_support_console
• Auth: None required (public API)
• Data shape: All timestamps are ISO-8601 with Z suffix. All enums are UPPER_SNAKE_CASE. All IDs are prefixed (see ID glossary below).

ID Prefix Glossary

Endpoint Catalog

Consumer / Business (Residential Fiber)

Mobile / Wireless

Enterprise

Meta

Data Models

Account (ACC-*)

account_id, name, service_area, status, tier
authentication: { last_login_status, account_recovery_status, last_login_at }

• status: Active | Suspended
• tier: always standard in this dataset
• authentication.last_login_status: SUCCESS | FAILURE
• authentication.account_recovery_status: "" (normal) | FAILURE

Ticket (TCK-*)

ticket_id, account_id, service_area, service_type, subscribed_mbps, status, issue_summary, created_at

• service_type: internet | video | voice
• subscribed_mbps: 100, 200, 300, 500, 750
• status: always OPEN in this dataset
• account_id: can be BAD-* (non-existent account → intake error)

Diagnostics (/api/diagnostics/<ticket_id>)

ticket_id, bandwidth_mbps, latency_ms, jitter_ms, root_causes[], started_at, completed_at

• root_causes (array of enum strings):
  • CONFIGURATION_DRIFT — account/profile out of sync
  • FIBER_DROP_DAMAGE — physical line damage
  • SIGNAL_LOSS — signal degradation
  • BACKBONE_CAPACITY — regional backbone overload
  • VOICE_PROFILE_STALE — voice config needs refresh
  • GENERATED_NOISE — non-deterministic filler (ignore for business logic)

Troubleshooting (/api/troubleshooting/<ticket_id>)

ticket_id, steps[], post_bandwidth_mbps, post_latency_ms, post_jitter_ms, started_at, completed_at

• steps (ordered array): PROFILE_REFRESH, PROVISIONING_SYNC, LINE_TEST, SIGNAL_REFRESH, BACKBONE_REROUTE_ATTEMPT, VOICE_PROFILE_REFRESH, GENERATED_CHECK

Outage (OUT-*)

outage_id, service_area, active, service_types[], impact_score, eta_hours, started_at

• active: boolean — only true outages affect service
• impact_score: 0.0–1.0 float
• service_types: array of affected service types

Case (CASE-*) — Mobile

case_id, customer_id, line_id, device_id, issue_type, customer_location, summary, opened_at

• issue_type: NO_SERVICE | MOBILE_DATA | SLOW_DATA | MMS
• customer_location: home | abroad

Line (LINE-*) — Mobile

line_id, customer_id, device_id, phone_number, plan_id, status, suspension_reason,
roaming_enabled, data_used_gb, contract_end_date

• status: Active | Suspended
• suspension_reason: "" | OVERDUE_BILL | CONTRACT_ENDED

Device (DEV-*) — Mobile

device_id, model, sim_status, signal_strength, speed_test, mobile_data_enabled,
airplane_mode, phone_roaming_enabled, wifi_calling_enabled, data_saver_mode,
vpn_connected, network_mode_preference, can_send_mms, mmsc_url_present,
messaging_permissions: { sms, storage }

• sim_status: active | missing | locked_pin
• signal_strength: none | good | excellent
• speed_test: no_connection | poor | fair | excellent
• network_mode_preference: 4g_5g_preferred | 3g_only

Plan (PLAN-*) — Mobile

plan_id, name, data_limit_gb, monthly_price_usd, data_refueling_price_per_gb

Bill (BILL-*)

bill_id, customer_id, amount_due_usd, due_date, status

• status: Paid | Overdue | Issued

Enterprise Account (ENT-*)

enterprise_account_id, name, tier, account_owner, finance_owner

• tier: Enterprise | Strategic

Enterprise Incident (INC-*)

incident_id, enterprise_account_id, product, severity, status, summary,
account_owner, engineering_owner, received_at

• severity: Critical | High | Medium
• product: monthly_export | dashboard_refresh | generated_product

Export Run (RUN-*)

run_id, enterprise_account_id, incident_id, run_date, status, exported_record_count, failure_code

• status: FAILED | SUCCEEDED
• failure_code: "" (success) | STALE_CREDENTIAL | STAGING_STORAGE_QUOTA

Enterprise Message (MSG-*)

message_id, author, body, channel, created_at

• channel: export-alerts-archive | account-escalations | data-platform | support

Enterprise SLA

enterprise_account_id, credit_trigger, monthly_export_credit_percent, executive_contact

Standard Operating Procedures

SOP-1: Triage a Consumer/Business Ticket

1. GET /api/tickets/<ticket_id> → extract account_id, service_area, service_type
2. GET /api/accounts/<account_id> → check status, auth
3. GET /api/outages?service_area=<service_area> → check for active outages covering service_type
4. GET /api/diagnostics/<ticket_id> → get root causes
5. GET /api/troubleshooting/<ticket_id> → get remediation steps

Decision tree:

• Account not found (BAD-* account_id) → Intake error — ticket references non-existent account
• Account status: "Suspended" → Account-level issue — billing/account hold
• Auth last_login_status: "FAILURE" or account_recovery_status: "FAILURE" → Authentication failure
• Active outage for service_area AND service_type in outage's service_types → Known outage (eta_hours, impact_score provide context)
• Otherwise → Technical issue — root_causes from diagnostics tell the story

SOP-2: Triage a Mobile Case

1. GET /api/cases/<case_id> → extract line_id, device_id, issue_type, customer_location
2. GET /api/lines/<line_id> → check status, roaming, data_used_gb, plan_id
3. GET /api/devices/<device_id> → check all device flags
4. GET /api/plans/<plan_id> → check data_limit_gb
5. If line Suspended: check suspension_reason, check /api/bills for bill status

Decision matrix by issue_type:

SOP-3: Investigate an Enterprise Incident

1. GET /api/enterprise/incidents/<incident_id> → product, severity, enterprise_account_id
2. GET /api/enterprise/export-runs?incident_id=<incident_id> → failure pattern
3. GET /api/enterprise/messages?query=<keyword> → human context (query by product, failure_code, account name)
4. GET /api/enterprise/sla/<enterprise_account_id> → credit terms
5. GET /api/enterprise/accounts → find account owner/finance owner

Failure code interpretation:

• STALE_CREDENTIAL → Credential rotation happened but scheduler still uses old secret. Fix: update credentials in scheduler.
• STAGING_STORAGE_QUOTA → Staging bucket full. Fix: increase quota or clear old data.
• Consecutive failures → Check SLA for credit trigger thresholds.

SLA credit patterns:

• Count consecutive failed runs (same failure_code, sequential dates)
• Match against credit_trigger text and monthly_export_credit_percent
• SLA credits are percentage of monthly bill

Common Pitfalls

1. Don't stop at the first finding. A ticket/case may have multiple contributing factors — check all related entities. For example, a "no service" case might have both a suspended line AND a missing SIM.

2. Outages only explain tickets in the same service_area. Always cross-reference ticket.service_area with outage.service_area, and verify ticket.service_type is in outage.service_types.

3. Not all outages are active. Only active: true outages affect current service. active: false outages are historical.

4. GENERATED_NOISE root causes are non-deterministic filler. Don't build business logic around them. Focus on the deterministic root causes: CONFIGURATION_DRIFT, FIBER_DROP_DAMAGE, SIGNAL_LOSS, BACKBONE_CAPACITY, VOICE_PROFILE_STALE.

5. GENERATED_CHECK troubleshooting steps are non-deterministic filler. Same rule — ignore for business logic.

6. Account ID BAD-* is a real pattern. Tickets can reference non-existent accounts. The BAD- prefix indicates an intake/registration error where the account was never created. Verify by checking /api/accounts/BAD-5403 (returns {"error": "not_found"}).

7. Roaming requires BOTH device AND line. A traveler abroad needs device.phone_roaming_enabled = true AND line.roaming_enabled = true. Check both.

8. Data cap is plan.data_limit_gb vs line.data_used_gb. Compare the plan limit with actual usage. data_used_gb > data_limit_gb = cap reached.

9. Contract expiration is a date comparison. contract_end_date is YYYY-MM-DD. If it's before the case opened_at date, the contract has ended (and the line should be suspended with CONTRACT_ENDED).

10. MMS failures have 3 distinct root causes. Check can_send_mms, then messaging_permissions.storage, then mmsc_url_present. They are independent checks.

11. Enterprise tiers matter for SLA. Strategic tier accounts may have higher credit percentages and tighter triggers than Enterprise tier.

12. Messages are best queried with keywords from the incident. Query by product name (export, dashboard), failure code (credential, quota), or account name (Asteri, Quanta).

Output Conventions

When producing a diagnosis, structure output as:

## Diagnosis for <ID>

**Root Cause:** <primary cause in plain English>
**Evidence:**
- <entity.field> = <value> → <what this means>
- ...

**Resolution:** <what was done or should be done>
**Metrics Impact:** <before → after for bandwidth/latency if applicable>

For multiple findings, list all with the most actionable/specific first. Distinguish between "this is the cause" vs "this is a contributing factor."

Data Generation Notes

• Some records contain "Generated" in names/summaries (e.g., "Generated Customer 01", "Generated support ticket"). These are synthetic filler records with non-deterministic values.
• Root causes labeled GENERATED_NOISE and troubleshooting steps labeled GENERATED_CHECK are non-deterministic — do not treat them as meaningful business signals.
• Generated incidents/products/incidents exist but the deterministic ones (INC-7301, INC-8301, INC-8402) contain the real business logic patterns.