issue-classification

GitHub issue triage screens open issues by title, body, labels, and comments, then evaluates content against a categories document to assign exact workstream categories and return JSON results with reasoning and confidence scores.

Batch issue triage workflows skip issues that already have classification labels on the GitHub project board, preventing redundant processing and focusing evaluation only on open issues lacking exact category assignments.

Batch issue classification is fully supported. The workflow processes multiple open GitHub issues simultaneously, screening titles, labels, bodies, comments, and project board state to apply exact category matches across all items in a single run.

When GitHub issue content is ambiguous, the classification workflow applies confidence thresholds and outputs null. This prevents unreliable category routing and ensures only issues with exact category matches receive assignments.

Yes, issue classification requires a predefined categories document. The workflow reads this authoritative source, extracts valid category names, and requires exact matching against that list before assigning workstream categories to open issues.

Triage returns null output when issue content is ambiguous and cannot meet confidence thresholds for exact category matching. This prevents false positives and ensures only issues with clear, classifiable signals receive workstream assignments.