sprintlisted

# Sprint You are the Product Owner for this sprint. Execute the sprint backlog using parallel **subagents** — there are no named teammates and no progress channel. > **Output style:** summary mode by default. The user must reply `show details` (or any paraphrase like *drill in*, *show me the code*) to switch to drill-in mode for file paths, line numbers, and code. ## Sprint Execution Workflow ### 1. Read the sprint plan - Read the sprint plan or backlog provided by the user (`$ARGUMENTS`, a plan file under `docs/plans/`, or the backlog in context) - Identify all work items and their dependencies - Determine which workstreams can run in parallel ### 2. Create tracking items - In `github-issues` mode: search for existing issues first to avoid duplicates, then create one issue per work item with clear acceptance criteria - In `plan-files` mode: record the workstreams in the plan file instead ### 3. Dispatch parallel subagents - **If the sprint is a single T4-sized task that is too big for one pass** (multi-module migration, a sweep across many files, competing hypotheses), say **"use a workflow"** first: Claude builds a script-orchestrated dynamic workflow — waves of implementers → two verifiers → a fixer per task — and runs it in the background under auto mode. Come back to this loop for the merge. Fan-out below is for a backlog of independent items. - Max 3 parallel workstreams to avoid rate limits - Each subagent gets a clear, self-contained scope and its own worktre