User-story review & ambiguity finder
The problem
Vague stories ('system should be fast', 'user can upload files') create defects later. Reviews depend on who has time.
Workflow
- Jira storydescription + AC
- LLM reviewrubric: clear, testable, complete
- Findingsgaps, questions, risks
- BA / POanswers in Jira
- Ready storytestable AC
Real-world example
Story: 'User can reset password.' The AI flags: no link expiry time, no rule for locked accounts, no message for unregistered email, no limit on reset attempts. The PO adds 4 acceptance criteria before sprint start.
Starter prompt
You are a senior QA reviewer. Review this user story against the rubric: clear, testable, complete,
consistent. List (1) ambiguous phrases, (2) missing acceptance criteria, (3) questions for the PO,
(4) top 3 risks. Do not invent business rules — ask.
STORY: {story}Tools
Jira + MCP or Jira API, any LLM
Human check
AI suggests; PO decides. Never auto-edit the story.
Measure success
Defects traced to unclear requirements per release; % stories with ≥1 finding.