icp-prompt-builder

You tune an ICP qualification prompt by iteratively scoring batches of 10 companies from list-builder outputs inside Claude Code Task sub-agents, providing user feedback on each batch until two consecutive rounds require zero corrections. The final tuned prompt is then saved for reuse.

The best way to qualify companies from list-building exports is to run them through an iterative AI scoring loop that applies your specific ICP criteria to the generated list, refining the evaluation rules until the automated qualification matches your expectations perfectly.

Yes, you can use this ICP prompt builder with Prospeo and Google Maps list exports, as well as outputs from DiscoLike and Blitz list-builders. It selects 10 test companies from these outputs to iteratively refine the qualification criteria.

No, you do not need an external API key to run AI company qualification. The prompt builder always uses Claude Code Task sub-agents natively for scoring and refining the ICP criteria without requiring any external API connections.

The iterative prompt coaching process works by building an initial qualification prompt, running it against a 10-company batch, gathering inline user feedback, and refining the criteria. This loop stops only when two consecutive scoring rounds have zero corrections.

You should run an ICP qualification prompt builder immediately after generating a sample list of companies and before scaling your outreach. It ensures your AI qualification criteria accurately match your client's ICP before processing larger lists.