What problem does it solve? List-building tools return raw companies without knowing whether they match your ideal customer profile, so scaling to thousands of unqualified companies wastes enrichment budget. This Skill tunes an AI qualification prompt on small batches before you pay to enrich the full list. ## Core Features & Use Cases - Iterative tuning loop: Scores batches of 10 companies via Claude Code Task sub-agents, collects user corrections, and refines the prompt until 2 consecutive rounds pass with zero corrections. - No external API keys: All scoring runs inside Claude Code using Task sub-agents, with no Anthropic or OpenAI SDK calls required during tuning. - Reusable output: Saves the converged prompt to a profile directory and records metadata in client-profile.yaml for applying at scale via scripts/score-batch.ts. - Use Case: After pulling 100 sample companies with a list-builder, run this Skill to converge on a qualification prompt in 3-5 rounds, then apply it to a 5,000-company list and keep only qualified entries with confidence >= 0.6. ## Quick Start Run the icp-prompt-builder skill to tune an ICP qualification prompt on 10 sample companies from my list-builder output.
