Scope Check

Sprints accumulate. A story added Tuesday because "it's quick" plus another Wednesday becomes Friday's overrun. This skill measures the drift from the original plan and surfaces a recommendation.

Read-only. Outputs a snapshot + cut recommendations.

Purpose / When to Run

Run when:

• The team feels overloaded mid-sprint
• A new story is being considered for an in-flight sprint
• Sprint status shows "At risk" and the cause is unclear
• A PM wants to confirm a feature has not silently expanded

Distinct from /sprint-status (current state) — this skill compares planned vs. actual scope.

Inputs

• Mode: sprint (compare current sprint to its initial plan) or feature <slug> (compare an epic to its initial scope)
• The relevant sprint plan or epic file
• Mid-Sprint Update sections (added by /sprint-plan update)
• All current stories
• The git history if available — but the skill does not require it

Outputs

• Printed snapshot. No file writes.

Phase 1: Resolve Mode

If no argument: ask

• Prompt: "Check sprint scope or feature scope?"
• Options: Active sprint, Specific feature (free text — epic slug)

Phase 2: Establish Baseline

For sprint mode:

• Read the active sprint plan
• Extract the original Sprint Backlog table (the one written when the plan was first created, before any ## Mid-Sprint Update section)
• Extract the original sprint goal and Success Looks Like

For feature mode:

• Read the epic's EPIC.md
• Extract the original Story Backlog (Section 13 if it was filled before stories were added; else infer from the original PRD's TR-IDs)
• Extract the original scope from EPIC Section 8 (Out of Scope)

Phase 3: Establish Current

Walk the same backlog now:

• For sprint mode: read the live story files for each story in the plan, plus any added via Mid-Sprint Updates
• For feature mode: list all stories under production/epics/<slug>/story-*.md

For each story now in scope, classify:

• In original plan — was there from day 1
• Added — appears in a Mid-Sprint Update section, or post-dates the original plan's create date
• Expanded — was in the plan but acceptance criteria have grown (count criteria; if the story file has more than the plan implied, mark as expanded)

For each story in the original plan:

• Carried — still here, on track
• Removed — not in the current list (deferred or killed)
• Stuck — still here but in Blocked or stalled In Progress for days

Phase 4: Quantify

Numbers to surface:

• Original story count
• Current story count
• Net additions (positive number = creep)
• Acceptance-criteria additions (sum of expansion across stories)
• Stories Blocked or stalled (drag, not creep, but distorts capacity)

For sprint mode, also:

• Day in sprint vs. % complete — if 60% time elapsed but 30% complete, the math may be unwinnable even without creep

Phase 5: Categorize Creep

For each Added story, classify:

• Linked to goal — directly serves the sprint goal or feature outcome (often justifiable)
• Adjacent — related but not on the critical path (cuttable)
• Opportunistic — "since we're touching this code anyway" (almost always cut)
• External — driven by external request mid-sprint (escalate to PM)

For each Expanded story, classify:

• Discovered necessity — turned out the original criteria were incomplete (acceptable; document)
• Polish — adds finish that wasn't in the original spec (cut or defer)
• Edge cases — surfaced during work; some essential, some defer-able (case by case)

Phase 6: Output

# Scope Check: <sprint NNN | feature <slug>>

## Drift summary
- Original scope: <N stories, M acceptance criteria>
- Current scope: <N+x stories, M+y acceptance criteria>
- Net creep: +<x> stories, +<y> criteria (<z>% bloat)

## Additions
- [LINKED] story-NNN: <title> — added on <date>, justified by <reason>
- [ADJACENT] story-NNN: <title> — added on <date>, recommend cut
- [OPPORTUNISTIC] story-NNN: <title> — recommend cut
- [EXTERNAL] story-NNN: <title> — escalate to PM

## Expansions
- story-NNN: <title> — original 3 criteria, now 6. <classification>.

## Removals (carry-out)
- story-NNN: <title> — moved out on <date>, going to <sprint or backlog>

## Drag (separate from creep)
- story-NNN: <title> — Blocked on ADR-0007 (Proposed)
- story-MMM: <title> — In Progress 4 days, last touched <date>

## Recommendation
**<Cut / Hold / Continue>**

If Cut: remove <list of N stories>, defer <list of M criteria>. New backlog count: <N-x>.
If Hold: scope is borderline. Drop <list of K opportunistic adds> if you want a buffer.
If Continue: scope drift is justified by goal-linkage.

## Recommended next steps
- Run `/sprint-plan update` to formally remove cut stories
- For drag items, escalate blockers (talk to architect / PM)
- Re-run `/sprint-status` after cuts to confirm burndown improves

Phase 7: Edge Cases

• No Mid-Sprint Update sections but story count is higher than the plan: stories were added without going through /sprint-plan update. Flag as a process gap.
• Expansion without addition: zero net adds, but criteria count up — quieter creep, often the dangerous kind.
• Plan was never tightly scoped: the original plan's Success Looks Like is vague. Flag — /scope-check cannot evaluate creep against a fuzzy baseline. Recommend tightening the next plan.

Quality Gates

• Numbers are reproducible — running twice on the same state gives the same numbers
• Recommendation is one of Cut / Hold / Continue — never "it depends"
• Read-only — no file writes
• Output is skimmable in 30 seconds

Examples

/scope-check

• Sprint 003: original 8 stories, 24 criteria. Current 11 stories, 33 criteria. Net +37%.
• 3 Added: 1 LINKED (justified), 1 OPPORTUNISTIC (cut), 1 EXTERNAL (escalate)
• 2 Expanded (1 polish, 1 edge case)
• 1 Drag (Blocked story)
• Recommendation: Cut. Remove the 1 OPPORTUNISTIC story and trim the 1 polish expansion. Net +1 story, manageable.

/scope-check feature paywall

• Epic paywall: original 12 stories from PRD. Current 17 stories.
• 5 Added: 4 ADJACENT (cuttable to v1.1), 1 LINKED
• Recommendation: Cut 4 adjacent stories. Defer to v1.1 epic.