sprint-plan

name

sprint-plan

description

Help the owner plan the next Duo sprint by harvesting candidates from tasks.md / active-sprint.md / session-log.md / roadmap.html, generating an interactive worksheet they prioritize (P0 / P1 / P2 / Skip), then synthesizing the result into a sprint commitment + active-sprint.md update. PROACTIVELY OFFER right after a cut lands or when the owner says "what's next?", "let's plan the next sprint", "let's triage", or signals end-of-chapter. Pairs with the `worksheet` skill (which it builds on) and `cut-version` (which it precedes).

sprint-plan skill — Duo

Why this skill exists. Sprint planning has been ad-hoc: scroll tasks.md, glance at active-sprint.md, brainstorm in chat, hope the right items make it into the next sprint. Captures: drift over time; P0 regressions sometimes get filed but not prioritized; "Other notes for next sprint" rows from the smoke walk get re-discovered weeks later. A worksheet-driven plan removes the scroll-and-remember step from the loop.

When to propose a sprint plan

Strong triggers (proactively offer):

1. A version cut just landed and the next sprint hasn't started.
2. The owner says any of: "what's next", "let's plan", "what should we prioritize", "triage this list", "what's in the next sprint", "let's get focused".
3. A smoke walk produced FAIL rows + "Other notes for next sprint" carry-overs that haven't been triaged yet.
4. After significant context compaction — a fresh sprint plan re-grounds the next session in concrete priorities.

Skip when:

• The current sprint is mid-flight (don't replan; finish what's committed).
• The owner asks for a one-off triage of a single item ("what should I do about BUG-074?" — answer in chat).

Procedure

1. Run the candidate gatherer

node .claude/skills/sprint-plan/gather.mjs <next-version>

gather.mjs writes the worksheet manifest to docs/dev/worksheets/sprint-plan-v<VERSION>.json by default. It pulls from:

• tasks.md — every entry whose status is 🆕 Filed / 🟡 Partial / ⏳ Open. Status badge appears in the per-item subtitle.
• docs/dev/active-sprint.md — the most recent walk's FAIL table (typically the highest-signal carry-overs) and the "Other notes for next sprint" table (already-triaged candidates).
• docs/roadmap.html — 🟡 in-progress stage cards (strategic context; only emits stages with a status-line emoji match).
• docs/dev/session-log.md — referenced in the worksheet's intro for the user to glance at while filling things out (parsing the prose log is unreliable; surfacing it as a link is better than emitting noise).

Items are deduped by ID; subtitle aggregates source provenance when the same item appears in multiple sources.

2. Skim the manifest before generating

The gatherer can produce noisy titles for FAIL rows whose original "Item" cell was just a bare ID — it falls back to the first sentence of the failure cell, which can include markdown bolding. Quickly read through docs/dev/worksheets/sprint-plan-v<VERSION>.json and hand-edit titles where the auto-derivation is awkward.

You can also cull obvious P3-or-lower noise here so the worksheet isn't 50 rows long. Aim for ~15-25 candidates.

3. Generate the worksheet HTML

node .claude/skills/worksheet/generate.mjs \
  docs/dev/worksheets/sprint-plan-v<VERSION>.json \
  docs/dev/worksheets/sprint-plan-v<VERSION>.html

4. Open in Duo

duo open docs/dev/worksheets/sprint-plan-v<VERSION>.html

Same hand-off as the smoke-walk skill — tell the user the worksheet title (e.g. "Sprint plan · v0.6.5") so they can find it in the tab strip if duo open didn't auto-focus it.

5. Hand off

Sprint plan worksheet Sprint plan · v0.6.5 is open as a browser tab. For each candidate: P0 if it must land in the next sprint, P1 if it would be great, P2 if nice-to-have, Skip to defer. Add notes for context, dependencies, or "this is just X rolled into Y." When you're done, hit Send to Claude at the bottom (or Copy results + paste back).

6. Parse the result + propose a sprint plan

When the user pastes back (or Send-to-Claude lands), parse the result block. Group items by priority. Then write a short proposal covering:

1. Sprint goal — one sentence summarizing the theme of the P0+P1 items. "Fix v0.6.4 cut-blockers + ship Split View Phase 3c browser-in-aux." Pull from the user's "Other notes" if they wrote one.
2. P0 commit — list the items, with rough order. These MUST land before the cut.
3. P1 stretch — what we'll attempt after P0 lands. Optional cut inclusion based on time.
4. P2 deferred — explicitly captured as "next sprint" items so they don't fall off the radar.
5. Open questions — anything the user flagged in notes that needs a decision before work starts.

7. Persist the plan

Write the proposal into docs/dev/active-sprint.md as a new section at the top, replacing the previous "next sprint" pointers in the existing file. Pattern:

## Sprint plan v<VERSION> — committed <date>

**Theme:** <one-line goal>

**P0 (cut blockers):**
- BUG-075 — Phase 3b chord regression (...)
- BUG-074 — Light-mode contrast (...)

**P1 (stretch):**
- ENH-083 — Move collapse-pane buttons (...)

**P2 (deferred to next sprint after this one):**
- ENH-080 — ⌘⇧A search open tabs

**Open questions:**
- ...

(see worksheet at docs/dev/worksheets/sprint-plan-v<VERSION>.html)

That section becomes the "what are we working on right now" answer for the duration of the sprint. The previous active-sprint.md content (the smoke walk results + carry-overs) moves down or gets pruned per the file's existing pruning rules.

8. Suggest next session start

Hand the user a one-line cue for the next session:

Next session, point me at sprint plan v and we'll start with the first P0 item ({first P0 item title}). I'll re-read active-sprint.md so we're grounded in the plan.

Files

• .claude/skills/sprint-plan/SKILL.md — this file.
• .claude/skills/sprint-plan/gather.mjs — candidate harvester. Reads tasks.md / active-sprint.md / roadmap.html and writes a worksheet manifest.
• docs/dev/worksheets/sprint-plan-v<VERSION>.json — generated manifest (gitignored).
• docs/dev/worksheets/sprint-plan-v<VERSION>.html — generated worksheet page (gitignored).
• docs/dev/active-sprint.md — destination for the committed plan.