Grid Anomaly Triage Agent Design Brief
Specify architecture, guardrails, and decision logic for an autonomous agent triaging SCADA alerts in power grids.
Use when engineering an autonomous agent to filter, correlate, and escalate power distribution anomalies. It aligns control room engineers and AI developers around deterministic safeguards.
Role: Principal Power Systems Automation Architect with twenty years in SCADA integrations and agentic AI systems.
Context
- Utility operator: {{utility_name}}
- Asset portfolio in scope: {{grid_asset_class}}
- Ingested data pipelines: {{telemetry_sources}}
- Real-time decision budget: {{latency_threshold_ms}}
- Human-in-the-loop governance: {{human_oversight_model}}
- Compliance benchmark: {{regulatory_framework}}
Task
Draft a comprehensive Grid Anomaly Triage Agent Design Brief that specifies agent behavior, contextual correlation rules, deterministic fallbacks, and escalation thresholds for {{utility_name}} across {{grid_asset_class}} assets.
Method
- Analyze {{grid_asset_class}} operational dynamics and baseline telemetry signatures from {{telemetry_sources}} to define baseline nominal states.
- Map raw telemetry noise patterns against actionable alarm topologies to construct the agent's noise-reduction pipeline.
- Formulate the agent's multi-step event correlation logic within the bounds of {{latency_threshold_ms}}.
- Define strict decision trees separating autonomous suppression, autonomous rerouting, and required escalations under {{regulatory_framework}}.
- Design the interaction model and state-handoff protocol under {{human_oversight_model}} when alert ambiguity exceeds confidence minimums.
- Specify deterministic fallback pathways if agent inferences stall, hallucinate, or lose upstream telemetry connectivity.
- Detail logging, audit trail retention, and explainability payloads required for post-incident root-cause investigation.
Constraints
- MUST enforce hard deterministic stops when critical safety thresholds are triggered, bypassing model inference.
- MUST NOT permit the autonomous agent to execute irreversible switching operations without {{human_oversight_model}} authorization.
- Analysis must strictly reflect the telemetry limitations of {{telemetry_sources}}.
- Keep technical prose concise, actionable, and free of vague generalizations.
Output format
Provide the brief structured in 4 distinct sections:
- Agent Architecture & Telemetry Pipeline (max 250 words)
- Anomaly Triage & Decision Logic (bulleted logic gates with confidence bands)
- Safety Guardrails & Human Handoff Protocols (3-5 explicit operational rules)
- Regulatory & Audit Verification Framework (table or structured list)
Self-review
- Confirm all telemetry inputs in {{telemetry_sources}} are accounted for in the triage flow.
- Verify that latency constraints ({{latency_threshold_ms}}) are reflected in the pipeline architecture.
- Check that human oversight requirements under {{regulatory_framework}} are strictly maintained.
Explicit role, a named task, and discrete steps the model can follow.
Background, inputs and variables the model needs before it starts.
Hard boundaries — what the model must and must not do.
A named, field-level shape for the response.
Ordered work items that force analysis before an answer.
Length and structure that travel across frontier models.
Signal density — instruction weight without padding.
Documented variables so the scaffold adapts to new inputs.
Quality bar, assumptions and behaviour when inputs are thin.
How much real usage the template has behind it.