Cold email list personalization: clean names and a one-line opener on every row

Loads the list with the script's own reader and checks the columns the skill needs: first name, company, city and state.

Lead CSV

Writes one small text file per chunk with just the row number, first name and company name, so each agent reads only what it needs.

personalize_leads.py

One subagent reads 20 lines and returns a nickname and a conversational company name for each, as JSON.

Subagent, personalize-leads skill

One subagent reads 20 lines and returns a nickname and a conversational company name for each, as JSON.

Subagent, personalize-leads skill

Joins both agents' JSON back onto the rows, undoes any HTML escaping, adds the lowercase state and renders the opener template on every row.

personalize_leads.py

Would fetch each company's homepage and look for 14 software signatures, adding columns that sort the list by likely spend.

personalize_leads.py, company websites

Runs the skill's mandatory sweep for surnames, cities and blank states, spot-reads the openers, then judges each flagged row: four held out, one kept.

personalize-leads skill

Writes the hand-off note: row counts, the by-state split, what was held and why, and the one-variable rule for the email body.

personalize-leads skill

Saves the 36 send-ready rows under the date-source-segment file name the staging convention asks for, with the held rows in their own file.

Send-ready CSV

Would export the named CSV to a new Google Sheet for a last look before it is loaded into the sending tool.

Google Sheets