project-orchestrator

Autonomous iteration planning based on evaluation failures is coordinated by reading project state, cycle reports, and failure buckets to select the next highest-value task. This ensures priorities are grounded in detected failure patterns and current repository metrics rather than intuition.

Prioritizing pipeline recovery and failure buckets in an autonomous project follows a deterministic decision order: restore broken pipelines first, then target the largest failure areas, and finally expand coverage or improve maintainability. This sequence ensures critical metrics-driven issues are resolved before optimization.

Generating actionable cycle briefs with owners and measurable goals involves reading the latest runtime JSON reports to produce a next-task recommendation. The output includes a designated owner, a specific goal, and acceptance criteria directly tied to the current repository state for validation.

Metrics-driven prioritization for long-running autonomous work operates across iterative cycles without manual intervention by continuously reading agent configuration and project state. It applies deterministic rules to evaluate dataset staleness and evaluation failures, ensuring the chosen next task always aligns with current project conditions.

Autonomous work cycles lack clear ownership and validation steps when they are not grounded in repo state and runtime reports. Applying an orchestrator resolves this by ensuring the chosen next task explicitly includes a clear owner, a measurable goal, and a defined validation step tied to project metrics.

Runtime JSON reports are required to start orchestrating autonomous skill cycles because the process depends on reading current cycle reports, metrics, and failure buckets. These inputs provide the necessary project state and evaluation signals to determine the next highest-value task accurately.