pm-classify-task

You can classify user feedback into PM task types by feeding messages into the system, which extracts intent keywords and maps them to feature, modification, bugfix, or refactor categories. It outputs a structured JSON patch with taskType and confidence for backlog refinement.

The best way to extract intent keywords from user stories is to analyze the messages for action-oriented terms that guide task scoping. The system identifies these keywords and links them to a specific PM task type to streamline requirement elicitation.

Yes, you can generate tasks automatically from user messages by producing a compact JSON patch. The patch contains fields like request.taskType, request.keywords, and notes, enabling downstream tooling to consume the structured output for automated task generation.

The four-way heuristic maps keywords to feature, modification, bugfix, or refactor by evaluating the detected intent within the user message. It assigns the most fitting PM task type and includes a recommended confidence value to indicate classification certainty.

No specific dependencies are required to run task classification on user messages. The system operates standalone without external components, allowing you to directly input user feedback and receive the classified task type patch output.

When auto-labeling tasks from user feedback, expect limitations in heuristic accuracy since classification relies on keyword matching rather than deep semantic analysis. Complex or ambiguous user messages may receive lower confidence values for their assigned task type.