← ClaudeAtlas

story-task-breakdownlisted

Turn a refined product story into a superpowers-style implementation plan — an ordered list of small, independently-verifiable tasks. Each task states a goal, concrete steps as checkboxes, and a verification. PO-readable but actionable; right-sized, TDD where it fits, never over-engineered.
itzikiusa/otto_os · ★ 0 · API & Backend · score 64
Install: claude install-skill itzikiusa/otto_os
# Story Task Breakdown You produce the **implementation plan** for a refined product story: a sequence of small, ordered, independently-verifiable tasks that a developer (human or agent) can execute one at a time, and that a Product Owner can read and track. This is *not* a design doc and *not* a re-statement of the story. It is the concrete plan of action: what to build, in what order, and how to know each step is done. > **Reference files live in `references/` and the output template is in `assets/` — both > sit alongside this SKILL.md. Consult them as you work:** > - `references/plan-format.md` — the EXACT heading/checkbox structure the UI parses, the > marker convention (`[ ]` / `[~]` / `[x]`), and how to phrase goals, steps, and verifies > - `assets/plan-template.md` — a filled example plan you can model your output on --- ## What "good" looks like (superpowers writing-plans style) A good plan is: - **Ordered.** Tasks run top to bottom. Earlier tasks unblock later ones. - **Small.** Each task is one coherent, reviewable change — typically minutes to a couple of hours, not "build the whole feature." If a task has more than ~7 steps, split it. - **Independently verifiable.** Every task ends with a concrete way to confirm it works (a test passing, a command's output, an observable behavior) — not "looks done." - **TDD where it fits.** When a task adds behavior that can be tested, write the failing test *first* as an early step, then make it pass. State this e