Prompt Library

Production-ready prompt templates for enterprise infrastructure operations. Each template includes token optimization metrics, model recommendations, and compliance metadata.

NOC

Fiber Fault Triage v2.1

Model: Claude Sonnet 4 (Local) Avg Tokens: 2,847 input / 1,203 output Latency: 4.2s

Template: Context: OTDR trace data, network topology JSON, historical fault log (last 30 days) | Task: Analyze trace for reflective events, correlate with physical plant data, classify fault type (bend, break, connector), estimate distance to fault ±0.1km. | Output: Structured JSON with fault_location, severity, recommended_dispatch_priority, confidence_score. | Guardrails: CPNI-safe (no customer identifiers in context), requires human approval for dispatch.

Node Health Monitor v1.4

Model: Local Llama 3.1 70B Avg Tokens: 1,455 input / 890 output Latency: 2.1s

Template: Context: SNMP telemetry from 1640/1670 nodes, temperature logs, power levels | Task: Identify degradation patterns (0.1dB loss trend), predict failure window, recommend maintenance window. | Output: Ticket_draft with priority_score, affected_circuits_list, suggested_action. | Optimization: Compressed context window using sliding 24h window, pruned redundant SNMP polls.

Compliance

CPNI Redaction Engine v3.0

Model: Claude Haiku 4 (High-throughput) Avg Tokens: 3,200 input / 45 output Latency: 0.8s

Template: Context: Raw customer communication (email, chat transcript, call log) | Task: Identify and redact CPNI (phone numbers, service addresses, network usage patterns) while preserving intent for customer service context. | Output: Redacted_text, cpnvi_flags_array, compliance_hash. | Compliance: Immutable audit log entry, zero-retention policy enforced.

FCC Audit Reporter v1.2

Model: Claude Sonnet 4 Avg Tokens: 5,600 input / 2,100 output Latency: 8.4s

Template: Context: Quarterly network usage logs, customer complaint records, outage reports | Task: Generate FCC Form 477 narrative summary, flag potential compliance gaps, suggest corrective actions. | Output: Formatted_report, compliance_score, risk_areas. | Note: Requires legal review before submission.

Field Ops

Truck Roll Optimizer v2.3

Model: Claude Sonnet 4 Avg Tokens: 1,890 input / 1,455 output Latency: 3.1s

Template: Context: GIS data, technician skill matrix, parts inventory, traffic patterns | Task: Route optimization for installation/repair dispatches, minimize drive time while matching skill requirements. | Output: Route_manifest, estimated_arrival_windows, required_parts_list. | Integration: Updates ServiceNow via MCP connector.

Installation Assistant v1.8

Model: Local Llama 3.1 8B (Edge device) Avg Tokens: 650 input / 420 output Latency: 0.4s

Template: Context: Customer premise equipment type, installation checklist, local code requirements | Task: Step-by-step guidance for fiber drop installation, splice verification, signal strength testing. | Output: Next_action, safety_warning_if_applicable, photo_documentation_prompt. | Deployment: Runs on ruggedized tablet, offline-capable.

Executive

Token Spend Analyzer v1.0

Model: Claude Sonnet 4 Avg Tokens: 8,400 input / 1,850 output Latency: 6.7s

Template: Context: API usage logs, department budgets, historical spend trends | Task: Identify anomalous token consumption, attribute spend to specific workflows, forecast monthly costs. | Output: Spend_report, anomaly_flags, optimization_recommendations, roi_calculation. | Audience: CFO, CIO, Department Heads.

AI Readiness Assessment v2.0

Model: Claude Sonnet 4 Avg Tokens: 4,200 input / 3,100 output Latency: 12.3s

Template: Context: Department process documentation, current tool stack, employee skill assessments | Task: Evaluate automation potential, rank use cases by ROI, generate phased implementation roadmap. | Output: Readiness_score, quick_wins_list, resource_requirements, risk_assessment. | Deliverable: Board-ready presentation deck outline.

Implementation Notes

All templates include:

• Version control: Git-based semantic versioning (major.minor.patch)

• Test suites: Regression tests against known-good outputs

• Cost tracking: Token counts logged per execution

• Access control: Department-scoped permissions via MCP

Token Optimization Standard: All prompts target <4K input context for standard operations. Extended context requires architectural review.

Compliance Standard: CPNI/PII templates undergo quarterly legal review. Audit logs retained for 7 years per FCC requirements.