| name | m-sdd-tasks |
| description | SDD phase 5 — normalize the plan's bite-sized steps into an ordered specs/<id>/tasks.md checklist. Triggers: sdd tasks, generate tasks sdd, task breakdown sdd, tasks from plan, sdd task list. |
SDD — Tasks
TL;DR — Gate on an approved plan (0) → extract the plan's checkbox tasks into specs/<id>/tasks.md with stable IDs (1) → mark done & route (2). Read <skills-root>/m-sdd/_shared.md first.
Pipeline: plan → tasks → analyze.
Phase 0 — Preflight & gate
- Read
_shared.md.
source <skills-root>/m-sdd/sdd-lib.sh; root="$(pwd)"; id="$(sdd_active_feature "$root")"sdd_require "$root" "$id" plan "approved" — the plan must be human-approved (not merely drafted) before it is broken into tasks.
Phase 1 — Derive tasks.md
sdd_set_phase "$root" "$id" tasks in_progress- Read
specs/$id/plan.md. m-writing-plans already emits bite-sized - [ ] steps; collect them into an ordered checklist in specs/$id/tasks.md (start from .sdd/templates/tasks-template.md). If the plan has no - [ ] steps, derive the tasks from its file-structure and Testing Strategy sections instead — and tell the user you did so.
- Assign stable IDs
T001, T002, … in dependency order. Keep each task independently testable. Do not invent work not in the plan.
- Traceability + tests: tag each task with the user-story / success-criterion IDs
it serves (
[US#], [SC-###]). Carry the plan's Testing Strategy into concrete
test tasks labelled the same way, so every story and SC-### has a verifying task.
UI test tasks name the design/<screen> they accept against. Whether missing
coverage is blocking is the constitution's call (default: expected, not blocking);
do not invent tests the plan's strategy didn't call for.
- Single source of truth: once
tasks.md exists, it is the canonical execution checklist — m-sdd-implement ticks boxes here, not in plan.md. To avoid two drifting copies, replace plan.md's ## Tasks body with a one-line pointer: See ./tasks.md (canonical).
Phase 2 — Record & route
- Self-check before marking done: verify every
US# and SC-### from spec.md appears on ≥1 task; fix or flag any gap first.
sdd_set_phase "$root" "$id" tasks done; route to m-sdd-analyze (then m-sdd-implement).
