Office DuckDB Read File

This skill is absorbed from the DuckDB official read-file workflow and adapted to ContextGo office analysis.

What it is for

Before writing queries, profile the file properly:

• what columns exist
• what row count looks like
• what the sample values suggest
• which reader or extension is needed

Use when

• The user asks what is inside a data file.
• You need to inspect a CSV, Parquet, JSON, Excel, SQLite, or similar file before deeper analysis.
• You want a file-aware profile before moving to SQL-style querying.

Do not use when

• The data model is already understood and the user is clearly asking for a query result.
• The task is centered on office documents rather than data files.

Workflow

Step 1: Resolve the file

Make sure the path or URL is real and identify:

• local or remote
• likely file type
• whether an extension may be needed

Step 2: Read with the right strategy

Use DuckDB's file readers or a read_any style macro mindset:

• CSV / TSV
• JSON
• Parquet
• Excel
• SQLite
• remote object storage when needed

If the generic path fails, switch to the exact reader instead of guessing blindly.

Step 3: Profile the file

Always try to produce:

• schema
• row count
• sample rows

This is the minimum useful profile.

Step 4: Turn the profile into next questions

After profiling, say:

• what the file most likely represents
• what quality issues are visible
• which query should come next

Output format

Return:

1. File profile

• file type
• likely content
• schema

2. Volume and sample

• row count
• sample notes

3. Data quality watchpoints

• missing values
• odd types
• suspicious fields

4. Recommended next query

• one or two SQL-style follow-ups

Use together with

• office-duckdb-query
• office-duckdb-install
• office-spreadsheet-analysis when the file is an Excel workbook acting as a dataset