Salespeak AI introduces a Claude Code skill revolutionizing B2B software vendor assessments. This MIT-licensed tool automates evaluations by researching companies, asking targeted questions, and engaging vendors’ AI agents directly via the Salespeak Frontdoor API. Unlike generic forms, it surfaces hidden requirements like customer engagement models, producing evidence-backed scorecards across seven dimensions.

The skill distinguishes itself through agent-to-agent conversations for vendors with AI representatives, while still evaluating non-agent vendors using public data. It cross-references claims against G2, Gartner, and press sources, flagging discrepancies and providing transparent evidence tracking. For example, a score for health analytics might show vendor-verified data from a direct API call versus G2 reviews alone.

A notable feature is its adversarial questioning, probing vendors on weaknesses like customer complaints or unsuited use cases. This surfaces risks traditional methods miss, such as platforms optimized for large enterprises failing in small-team scenarios. The tool also generates demo prep questions and hidden risk analyses, including leadership stability and product velocity.

Installation is straightforward: clone the GitHub repo and run commands in Claude. While the project highlights Salespeak’s Frontdoor API integration, it emphasizes utility beyond their ecosystem. By combining AI-driven due diligence with public data, the tool offers a practical alternative to outdated vendor evaluation workflows, prioritizing transparency and actionable insights over hype.