Spreadsheet & Data Wrangling Master

Files (3)

• README.md
• SKILL.md
• _meta.json

SKILL.md (first lines)

# Spreadsheet & Data Wrangling Master

> Turn messy data into clean insights, automated reports, and decision-ready dashboards. Platform-agnostic — works with CSV, Excel, Google Sheets, or any tabular format.

---

## Phase 1: Data Intake & Assessment

Before touching any data, assess what you have.

### Quick Health Check (score /20)

```yaml
data_intake:
  source: ""           # file path, URL, API, database, manual entry
  format: ""           # CSV, XLSX, TSV, JSON, clipboard paste
  rows: 0
  columns: 0
  file_size: ""
  encoding: ""         # UTF-8, Latin-1, Windows-1252, etc.
  delimiter: ""        # comma, tab, pipe, semicolon

  health_score:        # rate each 0-4, total /20
    completeness: 0    # 4=<1% missing, 3=<5%, 2=<15%, 1=<30%, 0=>30%
    consistency: 0     # 4=uniform types, 3=minor mixed, 2=significant mixed, 1=chaotic, 0=unusable
    accuracy: 0        # 4=verified, 3=plausible, 2=some outliers, 1=many errors, 0=untrustworthy
    freshness: 0       # 4=real-time, 3=<24h, 2=<7d, 1=<30d, 0=stale/unknown
    structure: 0       # 4=tidy (1 row=1 obs), 3=minor reshaping, 2=pivot needed, 1=multi-header, 0=freeform

  issues_found: []     # list every problem before fixing anything
```

### First 10 Questions to Ask

1. How many rows and columns?
2. What does each row represent? (one customer? one transaction? one day?)
3. Are there header rows? Multiple header rows? Merged cells?
4. What are the data types? (dates, currencies, percentages, IDs, free text)
5. How much is missing? Which columns?
6. Are there duplicates? By which key?
7. Is there a unique identifier column?
8. What date format? (MM/DD/YYYY vs DD/MM/YYYY vs YYYY-MM-DD vs mixed)
9. What currency/number format? (1,000.00 vs 1.000,00 vs 1000)
10. Where did this data come from and how often is it updated?

---

## Phase 2: Data Cleaning Decision Tree

### Step-by-Step Cleaning Protocol

```
START
  │
  ├─ Headers → Normalize (lowercase, snake_case, no spaces/special chars)
  │
  ├─ Duplicates?
  │   ├─ Exact duplicates → Remove, keep first
  │   ├─ Near-duplicates → Flag for review (fuzzy match on name + address)
  │   └─ Intentional duplicates → Leave (e.g., multiple orders same customer)
  │
  ├─ Missing Values?
  │   ├─ <5% of column → Fill (mean for numeric, mode for categorical, forward-fill for time series)
  │   ├─ 5-30% → Flag + fill with "UNKNOWN" or interpolate with justification
  │   ├─ >30% → Consider dropping column or flagging as unreliable
  │   └─ Entire row missing key fields → Remove with log
  │
  ├─ Data Types?
  │   ├─ Dates as text → Parse to date (try multiple formats, log failures)
  │   ├─ Numbers as text → Strip currency symbols, commas, whitespace, convert
  │   ├─ IDs/zips with leading zeros → Keep as text (NEVER convert to number)
  │   ├─ Phone numbers → Text, standardize format
  │   ├─ Mixed types in column → Split or coerce with error log
  │   └─ Boolean variants → Map (Yes/No/True/False/1/0/Y/N → consistent)
  │
  ├─ Outliers?
  │   ├─ Calculate IQR: Q1 - 1.5×IQR to Q3 + 1.5×IQR
  │   ├─ Business logic check (negative revenue? age 200? date in 2099?)
  │   ├─ Decide: fix (typo), cap (winsorize), remove, or keep with flag
  │   └─ ALWAYS log which outliers were modified and why
  │
  ├─ Standardization?
  │   ├─ Text case → Consistent (Title Case for names, UPPER for codes)
  │   ├─ Whitespace → Trim leading/trailing, collapse internal
  │   ├─ Categories → Map variants ("US"/"USA"/"United States" → "US")
  │   ├─ Dates → ISO 8601 (YYYY-MM-DD) internally
  │   ├─ Currency → Consistent symbol placement, decimal precision
  │   └─ Phone/email → Validate format
  │
  └─ Structural Issues?
      ├─ Multi-header rows → Flatten to single header
      ├─ Merged cells → Unmerge + fill down
      ├─ Pivot/crosstab → Unpivot to tidy format (1 row = 1 observation)
      ├─ Multiple tables in one sheet → Split to separate sheets/files
      └─ Metadata rows (totals, notes) → Separate from data rows
```

### Cleaning Log Template

```yaml
cleaning_log:
  date: "YYYY-MM-DD"
  source_file: ""
  rows_before: 0
  rows_after: 0
  actions:
    - action: "removed exact duplicates"
      rows_affected: 0
      key_columns: ["email"]
    - action: "filled missing values"
      column: "state"
      method: "mode"
      values_filled: 0
    - action: "removed outliers"
      column: "revenue"
      criteria: "negative values"
      rows_removed: 0
    - action: "standardized dates"
      column: "order_date"

From LeoYeAI/openclaw-master-skills, MIT.

More from LeoYeAI/openclaw-master-skills