PRD to GitHub Issues

This skill reads a product requirements document and decomposes it into a sequenced, estimable set of GitHub or Linear issues — one deliverable per ticket. It surfaces the work a PRD always omits: feature flags, migrations, instrumentation, rollback paths. Each issue comes out with a single-sentence scope, the context an engineer needs without re-reading the original spec, acceptance criteria written as observable behaviours you could demo, explicit blockers on other issues, and a size signal that flags which tickets need further breakdown before sprint planning.

You paste in your PRD — a Google Doc export, a Notion paste, or raw markdown — name your tracker, and describe your stack and team size. The skill asks four short clarifying questions before it starts, so the output reflects your constraints rather than generic defaults. If any section of the PRD is too vague to convert safely, the skill flags it explicitly and asks for clarification rather than converting ambiguity into a ticket.

A typical output excerpt for a payment-flow feature might look like: Issue 1 (S) — Add idempotency key column to payments table. Scope: schema migration only, no application logic. Acceptance criteria: column exists in production, backfill completes without downtime, rollback script tested. Blocks: Issues 3, 5. Issue 2 (M) — Implement idempotency check in charge endpoint. Depends on: Issue 1. Acceptance criteria: duplicate POST within 24 h returns cached response with original status code, not a new charge. The full output continues through every issue, grouped by milestone and ordered for parallel work where possible.