demo-write-user-stories

Demo: Writing User Stories (worked example)

Overview

This is a lean, from-scratch reference implementation of a real downstream consumer of CEP context packages — not a toy, and not part of core CEP. Given a feature description, it writes standard-form user stories, grounded in an approved context package's context_items when one is available, and follows the same discover/load/tag/reverse-index loop every consuming skill in this repo follows via CONSUMING-CONTEXT-PACKAGE.md.

It is deliberately minimal: no actor taxonomy, no Enabler/NFR/ Requirement-Note story types, no org-convention lookup, no scoring rubric. Any team that wants a production backlog tool should build one of those on top of this pattern — this skill exists to show the pattern itself, cleanly, end to end, and to be genuinely usable as-is for a quick, honestly-grounded first pass at a feature's user stories.

Scope note: this skill writes functional user stories only — plain "As a / I want / so that" statements. It does not distinguish Enabler stories, non-functional requirements, or requirement notes from functional stories; a production backlog tool should make that distinction, a lean reference example does not need to.

Inputs

1. Feature description (required) — one or two sentences describing the feature or behavior to write stories for, the way a developer would actually phrase a ticket.
2. Context package path (optional) — an explicit path to an approved contexts/<id>.yaml file to use instead of discovering one. If given, skip step 1's glob check below and treat that file as the "Found" package for step 2 onward, subject to the same existence/approval check (non-empty approved_by).

Steps

1. Discover or load a context package — follow .github/skills/ult-context-generate/CONSUMING-CONTEXT-PACKAGE.md steps 0–3, using the feature description as the input artifact for step 0's tag scan (it will usually carry no [Context: ...] tag for a bare ask — that's expected, fall through to step 1's glob check) and an explicit package path (if given) per "Inputs" above. Load context_items, decisions_log/decisions, aspects, and summary per step 3 if a package is found and approved.