The team behind Budna Marketplace has released Recommerce, a collection of portable agent skills designed to help AI assistants make smarter decisions when researching second-hand purchases. The repository, hosted on GitHub, introduces a structured framework for navigating the chaos of used goods, covering everything from discovery to final commitment. By treating a listing as "incomplete, changing evidence," the skills aim to distinguish between seller claims, product documentation, and facts that still need verification.

The Six Core Skills

Recommerce is not a single monolithic tool but a suite of six distinct skills: Buy, Kits, Inspect, Collect, Keep, and Second Look. The "Buy" skill generates a source-backed shortlist, while "Kits" assembles compatible bills of materials for multi-item setups. "Inspect" creates listing-specific cards that separate claims from checks, and "Collect" plans logistics like route and transport. "Keep" compares repair versus replacement costs, and "Second Look" reassesses earlier recommendations before a final commit.

Design Philosophy and Limitations

The project adheres to strict design principles, emphasizing that model specifications do not certify a particular used unit and that ratings do not establish condition. The skills are explicitly designed to keep assessment separate from transactions; they do not contact sellers, reserve items, bid, or purchase. This separation ensures that the AI acts as an analyst rather than an autonomous buyer, respecting the user's constraints without silently relaxing requirements or inventing availability.

Installation and Integration

Users can install the skills directly from GitHub using the Skills CLI with the command npx skills add budna-marketplace/recommerce. The installer allows for selective installation, such as adding only the "Buy" skill via --skill recommerce-buy. While the skills work independently with any evidence the host can access, an optional integration with Budna MCP is available for those needing specific tooling. The project is licensed under MIT.

Key Takeaways

  • Recommerce provides six modular skills for different stages of the second-hand buying process.
  • The framework strictly separates AI analysis from actual transactional actions like bidding or purchasing.
  • Installation is streamlined via the Skills CLI, allowing for global or project-specific setup.
  • The project relies on synthetic regression cases for development, not claiming general reliability.

The Bottom Line

Recommerce is a pragmatic step toward making AI agents useful in the messy real world of used goods, prioritizing evidence-based decision-making over blind trust in listings.