Troubleshooting Runbook Generator

# Task
Create a troubleshooting runbook. Use ONLY the provided system information.

# Inputs
- System context: {system_context}
- Symptoms: {symptoms}
- Environment: {environment}
- Constraints: {constraints}
- Known history: {known_history}

# Anti-hallucination rules
1. Do not invent system architectures, dependencies, or configurations.
2. If a diagnostic step requires access you're unsure exists, mark it [VERIFY ACCESS].
3. Clearly separate confirmed causes from hypotheses.

# Severity classification
Classify the issue based on symptoms:
- **SEV1** (Critical): Service down, data loss risk, customer-facing impact. Requires immediate response.
- **SEV2** (High): Degraded performance, partial outage, workaround available.
- **SEV3** (Medium): Non-urgent, limited impact, can wait for next business day.
State the classification and rationale.

# Runbook structure
1. **Issue summary**: Symptoms, affected systems, severity, initial hypothesis.
2. **Decision tree**: Step-by-step diagnostic flow. Use this format:
   - Step 1: Check [X]. If [condition A] → go to Step 2. If [condition B] → go to Step 4.
   - Each branch should narrow the cause. No step should take more than 5 minutes.
3. **Likely causes** (ranked by probability):
   - Cause, supporting evidence from symptoms, and fix.
4. **Rollback / safety checks**: Before applying any fix:
   - What to back up first
   - How to verify the fix didn't make things worse
   - Rollback procedure if the fix fails
5. **Escalation chain**: If still unresolved after the runbook:
   - Who to contact (role, not person name)
   - What information to provide them
   - SLA expectation per severity level
6. **Post-incident**: What to document after resolution.

# Output
Return the complete runbook with all 6 sections.

<context>
  <system_context>{system_context}</system_context>
  <symptoms>{symptoms}</symptoms>
  <environment>{environment}</environment>
  <constraints>{constraints}</constraints>
  <known_history>{known_history}</known_history>
</context>
<task>Create a troubleshooting runbook using ONLY provided system information.</task>
<instructions>
  No invented architectures. Mark uncertain access [VERIFY ACCESS]. Separate confirmed causes from hypotheses.

  Classify: SEV1 (critical) / SEV2 (high) / SEV3 (medium).

  Structure: Issue summary > Decision tree (step-by-step branching, each <5 min) > Likely causes (ranked) > Rollback/safety checks > Escalation chain (roles, info to provide, SLA) > Post-incident documentation.
</instructions>
<output>
  Return the complete runbook with all 6 sections.
</output>

System context: {system_context}
Symptoms: {symptoms}
Environment: {environment}
Constraints: {constraints}
Known history: {known_history}

Rules: only provided info, mark uncertain access [VERIFY ACCESS], separate confirmed from hypothesized causes.

Classify: SEV1/SEV2/SEV3.

Structure: Summary > Decision tree (branching steps, <5 min each) > Likely causes (ranked) > Rollback checks > Escalation (roles, info, SLA) > Post-incident.

Based on the entire content above, return the complete runbook.