Show HN: Ingot, evidence-gated optimization and version control for agent skills (github.com)

Ingot has been introduced as a local-first library and MCP server designed to enhance the management of agent skills through evidence-gated change control. This tool allows developers to version and promote skill instructions safely, ensuring that changes proposed by optimizers can only be activated after human approval. SkillOpt integration plays a crucial role by analyzing real agent performance, learning from historical traces, and generating evidence-backed proposals that allow thorough examination before any changes are adopted. This development is significant for the AI/ML community as it emphasizes safe experimentation and robust version control in agent skill management. By quarantining optimizer-generated proposals and requiring a detailed review process—complete with performance metrics and rollback capabilities—Ingot addresses potential risks associated with deploying untested modifications. The system not only serves current instructions efficiently but also encourages an iterative improvement approach, which is vital for the optimization of AI agents while minimizing downtime or unintended consequences. Through its user-friendly interface and local development environment, Ingot provides a powerful framework for individual developers to enhance their AI capabilities responsibly.