next-tasklisted

# /next-task — {{PROJECT_NAME}} sprint orchestrator (design-first) > **Announce on start:** open your reply with "Using /next-task to pick and dispatch the next sprint task." Find and execute the next task from the active sprint. Every task requires a Design Doc before implementation (design-first is non-negotiable — clear ports require upfront design). ## Token budget (MANDATORY) - Steps 1–6 (identify task): ≤15k tokens total. Use Grep + offset/limit; never full-Read files ≥200 lines. - Step 7 (create design doc): Read `docs/designs/_templates/DESIGN_TEMPLATE.md` ONLY if a fresh design doesn't already exist for this task. - Step 9 (dispatch): do not re-read CLAUDE.md — the harness auto-loads it; double-loading is waste. ## Steps 1. **Find the active sprint pointer** ``` Grep docs/spec/STATUS.md for the project row (or the active-sprint marker) ``` Extract: active sprint number, in-flight task ID, branch, last update. Fallback: `Grep` `docs/spec/backlog-index.md` for the active-sprint marker (e.g. `🚀`) → identify the active sprint pointer. 2. **Find the next un-started task in the sprint file** ``` Read docs/spec/sprints/sprint-S<N>.md with limit: 80 (header + task table only) Scan for the first row with status [ ] Not Started ``` - If a row is `[B] Blocked`, skip it. - If all rows are done → announce sprint complete, suggest `/retro`. - **Read the task's `Type:` slot** (feat / fix / refactor / chore / docs / spike / release).