# Prompt Design Solutions
Task 1: Quarterly Analysis Prompt

Explanation

Prompt::new("quarterly-analysis") .description("Comprehensive quarterly sales analysis with YoY comparison") .arguments(vec![ PromptArgument::new("quarter") .description("Quarter to analyze: Q1, Q2, Q3, or Q4") .required(true), PromptArgument::new("year") .description("Year (defaults to current)") .required(false), ]) .messages(vec![ PromptMessage::user(r#" Perform quarterly sales analysis for {{quarter}} {{year}}:

Step 1: Gather Context

• Read sales://schema to understand available data fields
• Read sales://regions to get the complete region list
• Note any schema changes that might affect comparisons

Step 2: Collect Current Quarter Data

• Use sales_query with date_range for {{quarter}} {{year}}
• Use sales_aggregate to calculate:
  • Total revenue
  • Units sold
  • Average order value
  • Customer count
• Break down by region using sales_aggregate with group_by="region"

Step 3: Collect Comparison Data

• Use sales_query with date_range for {{quarter}} of previous year
• Use sales_aggregate for same metrics
• Calculate year-over-year changes for each metric

Step 4: Identify Trends

• Compare regional performance: which regions grew/declined?
• Identify top 3 trends or anomalies
• Note any concerning patterns

Step 5: Generate Report Use report_generate with this structure:

Error Handling:

• If sales_query fails with RATE_LIMITED: wait and retry
• If data is missing for comparison period: note "No YoY data available"
• If any tool fails: report which step failed and what data is missing "#) ]) Prompt::new("bulk-update") .description("Safely update multiple customer records with preview and confirmation") .arguments(vec![ PromptArgument::new("update_type") .description("What to update: status, segment, or contact_info"), ]) .messages(vec![ PromptMessage::user(r#" Help me update customer records. This is a SENSITIVE operation.

Safety Protocol - Follow Exactly:

Phase 1: Understand the Request

• Ask what records should be updated (filter criteria)
• Ask what the new value should be
• Confirm the update_type matches: {{update_type}}

Phase 2: Preview (REQUIRED)

• Use sales_query to find matching records
• Display:
  • Total count of affected records
  • Sample of first 5 records with current values
  • If >100 records: STOP and ask user to narrow criteria

Phase 3: Confirmation (REQUIRED)

Present this summary:

Wait for explicit 'yes' response. Any other response = ABORT.

Phase 4: Execution (only after 'yes')

• Process in batches of 50 records
• After each batch, report: "Updated X of Y records..."
• If any error occurs: STOP and report what succeeded/failed

Phase 5: Verification

• Query updated records to confirm changes
• Report final summary:
  • Records successfully updated
  • Any failures
  • Rollback command if needed: bulk-update --rollback [batch_id] "#) ]) Prompt::new("sales-mode") .description("Enter sales data exploration mode with full context") .messages(vec![ PromptMessage::user(r#" Initialize a sales data exploration session.

Setup Phase:

1. Read sales://schema

   • List available tables and key fields
   • Note any date ranges or limitations

2. Read sales://regions

   • List all regions for reference
   • Note which have data

3. Read config://limits

   • Note current rate limits
   • Check query quotas remaining

Present Session Overview:

Session Rules:

For data questions:

• Use sales_query with reasonable LIMIT (default 100)
• Show result count and sample if large

For trend/aggregate questions:

• Use sales_aggregate instead of computing manually
• Explain what calculations were performed

For exports:

• Confirm before large exports (>1000 records)
• Use data_export and provide download info

For permission errors:

• Explain what's not accessible
• Suggest alternatives if possible

Maintain context across questions - reference previous results when relevant. "#) ]) :::

::: tests mode=local