Show HN: Orbit – A harness that makes AI coding agents prove their work (github.com)

Orbit has been introduced as an innovative harness designed to enhance the reliability and accountability of AI coding agents. This tool addresses common frustrations developers face, such as agents that fail to validate their outputs or leave no audit trail of changes. Orbit establishes a structured process for running coding agents, allowing them to complete tasks through a loop that includes selection, execution, validation, evaluation, and review. If an agent cannot substantiate its work, the process does not conclude, ensuring a rigorous check of results. The significance of Orbit for the AI/ML community lies in its capability to reduce the risks commonly associated with automated coding. It features dependency-aware task management, automated retry policies, and full telemetry to record each action taken by the agent. By integrating smoothly with various coding agents through an adapter-based system, Orbit allows for flexible use cases such as nightly bug triage, automated code health checks, and multi-agent handoffs. Its structured approach is set to improve the efficiency and trustworthiness of AI-driven coding workflows, ultimately paving the way for more robust software development processes.