roadmaplisted

# Roadmap ## Overview Standing inventory of everything a project still needs — features, pages, functions, integrations, infra — plus the dependency graph between them. `roadmap.json` is canonical; `ROADMAP.md` and `graph.{dot,svg}` are generated. A script computes the graph; **never compute "what's next" yourself.** This skill also **owns the hand-offs**: it decides when to grill, when to score, when one item needs a multi-PR blueprint, and how a wave of items gets executed. Those are gates with observable predicates below, not suggestions. ## When to use - "What should I work on next?" / "start the next thing" - A new feature, page, function or integration is mentioned — record it *first* - Project initialization, or an existing project with no roadmap - "What can run in parallel?" / batching work into waves - More than ~8 items competing for the same slot - An item finishes, or nothing reconciled in `RECONCILE_AFTER_DAYS` (7) days — ask `due`, never restate it **Not** for: how to build one item (`superplan`; `blueprint` for multi-PR), phase narrative (`PLAN.md`), vocabulary (`CONTEXT.md`, owned by `domain-modeling`), or where the whole run is going (`goal` — this skill orders items; a goal record such as `.claude/docs/goals/discovered-issues-closed.md` says which orderings count as progress). Capturing a defect found mid-run is this skill's job; **whether capturing it is the right answer at all** is `out-of-scope`'s — it scores fixing it now against routing it here