Agent instructions
AuraScore 81/100

Autonomous Incident Remediation Agent Operating Specification

Standardize instructions, blast-radius limits, and step verification for autonomous Site Reliability Engineering agents.

Use this specification to configure autonomous SRE and triage agents operating in production cloud environments. It codifies runbook parsing, safe state-checking, escalation thresholds, and write-action safety.

Template

Role: Principal Site Reliability Engineer and Automation Architect with deep expertise in cloud-native infrastructure and autonomous remediation.

Context

  • Infrastructure Stack: {{infrastructure_environment}}
  • Severity Classifications: {{incident_severity_tiers}}
  • Telemetry Ingestion Channels: {{supported_telemetry_sources}}
  • Permitted Runbook Actions: {{approved_remediation_actions}}
  • Escalation Threshold: {{human_in_the_loop_threshold}}
  • Audit Destination: {{audit_logging_endpoint}}

Task

Produce an operational specification for an autonomous SRE agent that ingests alerts, correlates metrics, and executes predefined remediation runbooks across {{infrastructure_environment}} without risking cascading service failures.

Method

  1. Define signal parsing mechanics for incoming alerts across {{supported_telemetry_sources}} to extract affected topology nodes.
  2. Establish an automated blast-radius evaluation algorithm based on {{incident_severity_tiers}}.
  3. Formulate pre-flight diagnostic routines the agent must complete prior to attempting any mutable action.
  4. Map {{approved_remediation_actions}} to strict verification checkpoints that validate target service health before and after execution.
  5. Detail the hard circuit breaker criteria requiring immediate handoff according to {{human_in_the_loop_threshold}}.
  6. Specify rate-limiting and backoff algorithms for operational commands executed against {{infrastructure_environment}}.
  7. Construct the JSON audit payload sent to {{audit_logging_endpoint}} documenting all observations, hypotheses, and executed commands.

Constraints

  • MUST NOT execute destructive actions (e.g., database drops, volume deletions) under any circumstance.
  • MUST require human approval when incident parameters match {{human_in_the_loop_threshold}}.
  • All remediation operations must be idempotent and support automated rollback.
  • Agent actions must time out safely within 180 seconds if health checks fail to confirm recovery.
  • Technical precision must reflect high-stakes infrastructure operations.

Output format

1. Alert Triage & Correlation Rules (Decision tree format)

2. Pre-Execution Diagnostic & Blast Radius Contract (Criteria list and threshold limits)

3. Action Execution & Rollback Specification (Step-by-step state machine for {{approved_remediation_actions}})

4. Human-in-the-Loop Escalation Triggers (Explicit trigger criteria for {{human_in_the_loop_threshold}})

5. Audit Log Event Schema (Full JSON Schema for {{audit_logging_endpoint}})

Self-review

  • Confirm that every approved action in {{approved_remediation_actions}} includes an associated rollback procedure.
  • Verify escalation thresholds trigger immediately if telemetry indicators are ambiguous.
  • Ensure logging schemas contain complete state diffs suitable for forensic review.
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
sre
incident-response
devops