.png&w=384&q=75)
Will Leatherman
Founder, Catalyst GTM Agency
Ask a chat for 50 B2B leads and it hands you 50 names with no reason behind any of them. That list costs more than an afternoon. Fullcast and Ebsta looked at 440K opportunities and found only 23% of pipeline was a strong ICP fit, and those deals were 8x more efficient to win.
This is the starter kit from Build a Customer-Finding Agent in Claude Cowork. It is one folder with three agents in it, and each one has one job. The finder searches public pages for companies that might match your ICP. The scorer checks each one against five checks and sorts the best fits to the top. The opener writes why each strong fit matches and one first line that reacts to something they published, with the quote and the link next to it.
Every fact in the final list carries a link, and anything the agent can't confirm scores zero. Checking a row takes seconds. Sending, connecting and adding anyone to a CRM always wait for you.
Every file ships finished, so you can compare yours against ours at each step.
What's in the download:
prompts/sharpen-icp.md, the prompt that interviews you one question at a time and rewrites your ICP
icp.md, a finished ICP with five checks, a trigger, who to leave out and the title that signs
CLAUDE.md, the house rules every agent reads first, including the six that keep the list honest
Three finished agent files for the finder, the scorer and the opener
A template for the ready list and an example with placeholder companies, so you see the shape before your first run
BUILD_STEPS.md, the six steps in order with every prompt, including the one that makes it a weekly scheduled task
PREREQUISITES.md and a README, so you know what to have open before step 1
You need the Claude desktop app with Cowork on your account and web search turned on. Scheduled runs read files on your computer, so it has to be awake with Claude open.
Built by Will Leatherman and the team at Catalyst, who've worked with 150+ B2B companies on go-to-market and AI search.
Free
Turn one sentence about your best customer into a weekly list of scored companies, each with a reason and a first line.