Agent instructions
AuraScore 81/100

Principal Reliability Architect Specification: SRE Autonomous Incident Agent

Architect a production-grade autonomous SRE triage and remediation agent specification with safety sandboxes, telemetry checks, and escalation guardrails.

Use this template when designing autonomous incident response agents for cloud services. It creates a robust system specification covering state machines, remediation safety, and on-call escalation.

Template

Role: Principal Reliability & Automation Architect with 15+ years designing self-healing distributed cloud infrastructure.

Context

  • Target deployment ecosystem: {{target_service_ecosystem}}
  • Telemetry and log ingestion pipeline: {{telemetry_source_stack}}
  • Allowlisted automated actions: {{permitted_remediation_actions}}
  • Escalation trigger thresholds: {{human_escalation_criteria}}
  • Target classification rubric: {{incident_severity_framework}}
  • Diagnostic query boundary: {{observability_retention_window}}

Task

Author a production-grade system prompt and deterministic operational specification for an autonomous Site Reliability Engineering (SRE) incident triage agent, establishing diagnostic routines, action safety sandboxes, and zero-downtime escalation boundaries.

Method

  1. Define the agent's core state machine, formalizing transitions across Detection, Corroboration, Root-Cause Hypothesis, Safe Mitigation, and Escalation phases.
  2. Map incoming alerts from {{telemetry_source_stack}} against {{incident_severity_framework}} to classify alert veracity and blast radius within {{target_service_ecosystem}}.
  3. Specify automated diagnostic queries constrained by {{observability_retention_window}} to isolate anomaly signatures without overloading observability backends.
  4. Construct a strict parameter-validation matrix for all executable commands in {{permitted_remediation_actions}}, defining pre-flight safety checks and dry-run expectations.
  5. Formulate human-in-the-loop interruption protocols based on {{human_escalation_criteria}}, defining exact payload structures for on-call engineer paging.
  6. Design idempotency and rollback verification loops to confirm whether automated interventions successfully mitigated the degraded state.
  7. Generate concrete prompt directives, negative instruction guardrails, and structured telemetry generation formats for agent self-logging.

Constraints

  • MUST enforce dry-run validation before executing any item in {{permitted_remediation_actions}}.
  • MUST NOT grant direct database write permissions, schema modification access, or persistent privilege escalation.
  • MUST invoke immediate human handoff upon encountering ambiguous telemetry matching {{human_escalation_criteria}}.
  • State transition schemas must output strictly valid JSON representations alongside descriptive system guidance.
  • All remediation execution steps must include explicit timeout and rollback definitions.

Output format

Provide a comprehensive markdown specification structured into these exact sections:

  1. Executive Agent Directive (System prompt persona, core objective, and behavior boundaries, max 250 words)
  2. Operational State Machine (Detailed step-by-step state transition logic with trigger conditions)
  3. Tool Execution & Action Safety Matrix (Table: Tool Name, Permitted Parameters, Safety Pre-checks, Rollback Action)
  4. Escalation & Paging Payload Schema (JSON Schema for alert payloads routed to human responders)
  5. Incident Post-Mortem Logging Contract (Structured logging schema for agent-initiated actions)

Self-review

  • Verify all variables ({{target_service_ecosystem}}, {{telemetry_source_stack}}, {{permitted_remediation_actions}}, {{human_escalation_criteria}}, {{incident_severity_framework}}, {{observability_retention_window}}) are integrated into system logic.
  • Confirm all remediation actions require pre-flight checks and rollback routines.
  • Verify escalation conditions prevent destructive agent loops and deadlocks.
AuraScore breakdown
81/100Provisional
Instruction clarity15/15 · Strong

Explicit role, a named task, and discrete steps the model can follow.

Context architecture12/12 · Strong

Background, inputs and variables the model needs before it starts.

Constraint engineering12/12 · Strong

Hard boundaries — what the model must and must not do.

Output specification6/14 · Thin

A named, field-level shape for the response.

Reasoning structure10/10 · Strong

Ordered work items that force analysis before an answer.

Model compatibility10/10 · Strong

Length and structure that travel across frontier models.

Token efficiency5/10 · Thin

Signal density — instruction weight without padding.

Reusability7/7 · Strong

Documented variables so the scaffold adapts to new inputs.

Robustness3/5 · Adequate

Quality bar, assumptions and behaviour when inputs are thin.

Observed performance1/5 · Thin

How much real usage the template has behind it.

ai-agents
agents-instructions
technology-software
agent instructions
sre
incident response