Skip to content

CLAUDE.md

Latest commit

 

History

History
126 lines (100 loc) · 5.51 KB

CLAUDE.md

File metadata and controls

126 lines (100 loc) · 5.51 KB

10x-Analyst Plugin — Agentic Analysis Swarm

This is a Claude Code plugin by 10x.in — a multi-agent swarm for end-to-end data analysis automation.

Plugin Commands

  • /10x-analyst:analyze — Full agentic pipeline (ingest → clean → analyze → visualize → report → dashboard)
  • /10x-analyst:profile — Data profiling and quality assessment only
  • /10x-analyst:clean — Data cleaning and transformation only
  • /10x-analyst:query — Ask natural language questions about your data
  • /10x-analyst:visualize — Generate charts and visualizations from data
  • /10x-analyst:report — Generate a comprehensive Markdown analysis report
  • /10x-analyst:dashboard — Build a standalone interactive HTML dashboard

Agent Swarm Architecture

The plugin coordinates 5 specialist agents in a pipeline:

User Request
     │
     ▼
┌─────────────────┐
│  ORCHESTRATOR    │  ← CLAUDE.md (this file) routes to agents
│  (Command Router)│
└────────┬────────┘
         │
    ┌────┼────┬──────────┬────────────┐
    ▼    ▼    ▼          ▼            ▼
┌──────┐┌──────┐┌──────────┐┌────────┐┌──────────┐
│ Data ││Stats ││Visualizer││Reporter││Strategist│
│Engine││ician ││          ││        ││          │
└──┬───┘└──┬───┘└────┬─────┘└───┬────┘└────┬─────┘
   │       │         │          │           │
   ▼       ▼         ▼          ▼           ▼
 Clean   EDA &    Charts &   Markdown    Business
 Data    Stats    Dashboard   Report     Actions

Agent Responsibilities

Agent File Role Delegates To
Data Engineer agents/data-engineer.md Ingest, profile, clean, transform data files Scripts: profiler.py, data_cleaner.py
Statistician agents/statistician.md EDA, correlations, distributions, statistical tests, RFM —
Visualizer agents/visualizer.md Matplotlib/seaborn charts, Chart.js HTML dashboards Scripts: chart_generator.py, dashboard_template.py
Reporter agents/reporter.md Assemble findings into structured Markdown report References: analysis-patterns.md
Strategist agents/strategist.md Interpret findings, generate business recommendations & action items —

Pipeline Flow by Command

Command Agents Used (in order)
:analyze Data Engineer → Statistician → Visualizer → Reporter → Strategist
:profile Data Engineer only
:clean Data Engineer only
:query Data Engineer → Statistician → Strategist
:visualize Data Engineer → Visualizer
:report Data Engineer → Statistician → Reporter → Strategist
:dashboard Data Engineer → Statistician → Visualizer

Path Resolution

Every command takes a dataset name as argument (e.g., shopify-data). The orchestrator resolves paths:

  • Input: input/<dataset-name>/ — where data files are read from
  • Output: output/<dataset-name>/ — where all artifacts are written

All paths are relative to the 10x-analyst/ plugin root. Never read or write outside the plugin directory.

Supported Data Sources

  • CSV (.csv) — pandas read_csv
  • Excel (.xlsx, .xls) — pandas read_excel via openpyxl/xlrd
  • JSON (.json) — pandas read_json / json_normalize

Input / Output Directories

All data to analyze must be placed inside input/ (in a subfolder per dataset). All artifacts are written to output/ (auto-creates a subfolder per analysis run).

10x-analyst/
├── input/                         # PUT YOUR DATA HERE
│   └── shopify-data/              # Example dataset (included)
│       ├── customers.csv
│       ├── orders.csv
│       ├── order_items.csv
│       ├── products.csv
│       └── price_changes.csv
│
└── output/                        # ALL RESULTS GO HERE
    └── shopify-data/              # Auto-created per dataset
        ├── report.md
        ├── dashboard.html
        ├── data-profile.md
        ├── cleaning-log.md
        ├── insights.json
        ├── cleaned-data/
        └── charts/

Key Directories

  • agents/ — Specialist subagent definitions (5 agents)
  • skills/ — Plugin slash commands (7 commands)
  • references/ — Analysis patterns, chart styles, data quality standards
  • scripts/ — Reusable Python utilities for profiling, cleaning, charting, dashboards

Model Strategy

  • Opus: :analyze full pipeline, :query complex questions (maximum reasoning)
  • Sonnet: :report, :dashboard generation (balanced quality/speed)
  • Haiku: :profile, :clean, :visualize (token-efficient mechanical tasks)

Demo Dataset

A Shopify e-commerce dataset is included at input/shopify-data/:

  • customers.csv — Customer profiles
  • orders.csv — Order transactions
  • order_items.csv — Line items per order
  • products.csv — Product catalog
  • price_changes.csv — Historical price changes

Quick start: /10x-analyst:analyze shopify-data

This reads from input/shopify-data/ and writes all results to output/shopify-data/.