Field Extraction Cheatsheet

Use this skill when cleaning CSVs, extracting emails/addresses/dates, or flattening JSON for tabular output.

Field Extraction Cheatsheet

CSV hygiene

• Always detect delimiter; don't assume comma
• Normalize headers: lower().strip().replace(" ", "_")
• Dedup keys: choose stable business keys, not row index
• Quote fields containing commas; use csv module, not string split

Structured extraction patterns

• Email: \b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}\b
• ETH address: ^0x[a-fA-F0-9]{40}$
• ISO date: \d{4}-\d{2}-\d{2}
• USD amount: \$?\d{1,3}(,\d{3})*(\.\d{2})?

Agent deliverable template

1. One-line summary of what was extracted
2. Schema of output fields
3. Sample rows (3–5)
4. Edge cases / null handling
5. How to re-run

JSON → tabular

• Prefer flat records; explode nested lists into child tables
• Keep ids for joins
• Validate types before write

Quality checks

• Row count in == out (or explain drops)
• Null rate per column
• Unique key collisions

Install

Copy this SKILL.md into your agent skills folder and invoke when doing structured extraction tasks.