Skip to main content

Finance Review Orchestrator

Purpose​

Execute the complete financial review pipeline using specialized self-validating agents.

Pipeline Stages​

/finance-review <month> <csv_files...>
    │
    ├─> 1. finance-csv-normalizer (validates: structure + balance)
    ├─> 2. finance-transaction-categorizer (validates: CSV structure)
    ├─> 3. finance-account-merger (validates: merged output)
    ├─> 4. finance-graph-generator (validates: PNG generation)
    └─> 5. finance-dashboard-generator (validates: HTML structure)

Arguments​

  • $1 - Month identifier (jan, feb, mar, etc.)
  • $2+ - One or more raw CSV files to process

Usage​

# Process January data
/finance-review jan raw_checkings_jan.csv raw_savings_jan.csv

# Process single file
/finance-review mar transactions.csv

Workflow​

Stage 1: Normalize​

For each input CSV file:

  1. Spawn finance-csv-normalizer agent
  2. Convert bank-specific format to standard schema
  3. Validate column structure and balance consistency

Stage 2: Categorize​

For each normalized CSV:

  1. Spawn finance-transaction-categorizer agent
  2. Auto-categorize transactions
  3. Validate CSV structure preserved

Stage 3: Merge​

  1. Spawn finance-account-merger agent
  2. Combine all normalized CSVs into single file
  3. Sort by date chronologically
  4. Validate merged output

Stage 4: Graph​

  1. Spawn finance-graph-generator agent
  2. Generate 8 financial insight visualizations
  3. Validate all graphs created

Stage 5: Dashboard​

  1. Spawn finance-dashboard-generator agent
  2. Create interactive HTML report
  3. Embed graphs and summary statistics
  4. Validate HTML structure

Output Structure​

data/<dataset>/mock_dataset_<month>_1st_2026/
├── raw_checkings.csv           # Original input (copied)
├── raw_savings.csv             # Original input (copied)
├── normalized_checkings.csv    # Standardized + categorized
├── normalized_savings.csv      # Standardized + categorized
├── agentic_merged_transactions.csv  # All accounts merged
├── index.html                  # Monthly dashboard
└── assets/
    ├── plot_01_spending_by_category_pie.png
    ├── plot_02_daily_spending_trend.png
    ├── plot_03_income_vs_expenses.png
    ├── plot_04_top_merchants.png
    ├── plot_05_category_over_time.png
    ├── plot_06_running_balance.png
    ├── plot_07_spending_distribution.png
    └── plot_08_spending_by_weekday.png

Cumulative Data​

After processing, also updates:

data/<dataset>/agentic_cumulative_dataset_2026.csv  # Year-to-date
data/<dataset>/index.html                            # Yearly dashboard

Parallel Execution​

Stages 1-2 can run in parallel for multiple files:

├─> normalize (checkings.csv) ─┐
├─> normalize (savings.csv) ───┼─> merge ─> graph ─> dashboard
└─> normalize (credit.csv) ────┘

Error Handling​

Each stage has self-validating agents that:

  1. Detect validation failures
  2. Receive specific error messages
  3. Self-correct and retry
  4. Block completion until valid
  • Agents: finance-csv-normalizer, finance-transaction-categorizer, finance-account-merger, finance-graph-generator, finance-dashboard-generator
  • Skill: self-validating-agent-patterns
  • Hooks: csv-structure-validator, csv-balance-validator, graph-file-validator, html-structure-validator