Summary

Enhanced the PRD parser to correctly handle user story descriptions in multiple LLM output formats, and updated the PRD generation guidance to specify plain text descriptions without bold prefixes.

Changes

PRD Generation Guidance (src/chat/engine.ts)

• Updated PRD_COMPATIBILITY_GUIDANCE to clarify that user story descriptions should be plain text on the next line following the format: "As a user, I want to ... so that ..."
• Removed the **Description:** bold prefix from the example format
• Added explicit note: "IMPORTANT: User story descriptions must be plain text (no Description: prefix)"

Description Extraction Logic (src/prd/parser.ts)

• Added DESCRIPTION_STOP_PATTERN regex to identify known metadata fields (**Acceptance Criteria:**, **Priority:**, **Depends on:**, **Labels:**, **Notes:**) that should terminate description extraction
• Added DESCRIPTION_LABEL_PATTERN regex to detect and strip the **Description:** label prefix that some LLMs generate
• Enhanced extractStoryDescription() function to:
  • Replace the generic line.startsWith('**') check with the more precise DESCRIPTION_STOP_PATTERN match to avoid false positives
  • Strip the **Description:** label prefix from lines before processing
  • Remove any remaining bold emphasis markers (e.g., **As a** → As a) from the final description text
• Updated JSDoc to document the three LLM output formats now supported:
  • Plain text: "As a user, I want..."
  • Bold-label: "Description: As a user, I want..."
  • Bold-keyword: "As a user I want to... So that..."

Implementation Details

The parser now gracefully handles inconsistent LLM formatting by:

1. Detecting and stripping known metadata field prefixes to prevent premature termination
2. Removing the **Description:** label if present
3. Stripping any remaining bold formatting used for emphasis within the description text
4. Joining lines and trimming whitespace to produce clean, plain text descriptions

This makes the parser more robust to variations in LLM output while encouraging the generation of properly formatted descriptions going forward.

Summary by CodeRabbit

Release Notes

• Improvements

  • Updated user story description formatting to require plain text without markdown prefixes for better consistency and clarity.
  • Enhanced description parsing to handle multiple formatting variations and metadata boundaries more robustly.

• Tests

  • Added comprehensive test coverage for PRD description parsing and formatting guidance.

_{✏️ Tip: You can customize this high-level summary in your review settings.}