Show HN: Receipts, agent skills that make PRDs and metrics cite their evidence (github.com)

A new tool, "Receipts," has been introduced to enhance the reliability of product documentation in AI/ML environments by requiring that claims in product requirements documents (PRDs), metrics, and launch copies are substantiated by evidence. Developed by Ali Raza, this set of six agent skills—compatible with Claude Code, Codex, and Cursor—ensures that any claims made about AI systems' effectiveness or reliability are either sourced, measured, or clearly noted as unsupported. By doing so, Receipts promotes accountability within product management, compelling teams to provide clarity on what specific metrics or outcomes genuinely demonstrate before launching products. The significance of Receipts lies in its potential to streamline the documentation process while enhancing the integrity of AI products, addressing a common issue where vague claims, such as "97% reliable," proliferate without substantive backing. Each skill serves a specific purpose, such as providing readiness gates for AI features or filtering through A/B test plans. This approach not only cultivates better communication in product development but also helps consumers and stakeholders understand the limitations of AI technologies, consequently leading to more informed decisions. As AI/ML continues to evolve, tools like Receipts are crucial for fostering transparency and trust in AI solutions.