dx-roadmap

Turn a researched, framed effort into an ordered sequence of vertical slices at context/efforts/<effort-id>/roadmap.md. Each slice is a tracer bullet — end-to-end and demoable, not a horizontal layer — and maps to exactly one child change. This skill decomposes; it does not create the child changes (that is dx-new).

Guard. Resolve <effort-id> under context/efforts/. If it is missing, tell the user to run /dx-new first. If the path is under context/archive/, refuse — an archived effort is done. If roadmap.md already exists, show it and ask before overwriting.

1 — Gather what upstream settled

Read effort.md (note its ## Goal). Then read the effort's shared upstream as context: every research/<topic>.md, and frame.md and brainstorm.md if present. A brainstorm.md reached this effort by passing the bundling test — its ## Conclusion & route names the independently shippable capabilities it found, which is raw material for the slices below. Read foundation/glossary.md for naming (a one-line habit — no section). These are decisions already made; slice within them, don't re-litigate them.

2 — Draft candidate slices

Decompose the goal into candidate slices — don't order or write them yet, just name and scope each:

• Feature effort — each slice cuts end-to-end through every layer it touches (schema → api → ui), narrow but complete. Not "all the schema, then all the api."
• Refactor effort (from dx-refactor-discover) — one slice per module deepening.
• A slice must be able to ship alone. A candidate that only hardens or extends another is folded into that slice as a phase — tell the user which and why.
• The lead slice is the tracer: one happy path and one negative test, end to end.
• For each candidate, note its dependencies: which other candidates (if any) it needs in place first.