Jira Tickets

PRD in, tickets out. Each ticket should be pick-up-able by a dev who only reads that ticket.

Step 1: Gather inputs

• The PRD (usually docs/prd/*.md). If there isn't one and the request is fuzzy, suggest prd-writer first.
• The repo map, for which repo owns what.
• An impact-analysis report if one exists. If the feature touches shared models or APIs and there's no report, run impact-analysis first or note the risk on the tickets.
• Ask once: Jira project key, and whether the team uses story points or hours. Default to story points (1, 2, 3, 5, 8).

Step 2: Slice

• One epic per feature. Tickets live under it.
• Tickets are per repo. A change that spans backend and frontend is at least two tickets with a dependency between them.
• Slice vertically where possible: a ticket delivers something testable (endpoint + tests), not "write models" then "write views" then "write tests".
• Separate tickets for anything with its own deploy risk: migrations on big tables, backfills, contract changes to shared APIs.
• Anything over 8 points gets split.