definition

Definition Skill

Entry condition check

Before asking anything, verify:

1. Discovery artefact exists at artefacts/[feature]/discovery.md - status "Approved"
2. Benefit-metric artefact exists at artefacts/[feature]/benefit-metric.md
3. Benefit-metric artefact contains at least one metric with a defined target

Also check for reference materials at artefacts/[feature]/reference/. If present, scan for technical designs, scoping documents, or programme-level constraints that should shape story decomposition and are not yet reflected in the discovery or benefit-metric artefacts. Surface any that are relevant before choosing a slicing strategy.

If not met (entry condition):

❌ Entry condition not met [Specific issue - e.g. "Benefit-metric artefact not found. Run /benefit-metric first."]

Run /workflow to see the current pipeline state.

---

Step 1 - Confirm the feature scope

State what was found first:

Discovery loaded: [feature name] MVP scope items found: [n] Benefit metrics found: [n metrics with targets] Personas identified: [list]

Ready to decompose this into epics and stories? Reply: yes - or name a specific scope item to focus on first

---

Step 1.5 - Architecture constraints scan

Before choosing a slicing strategy, check for constraints that should shape how stories are written.

Read .github/architecture-guardrails.md if it exists.

If found, surface relevant constraints before decomposing:

Architecture constraints relevant to this feature:

• [e.g. ADR-002: all data access must go via the API layer - no direct DB calls from UI]
• [e.g. Mandatory constraint: all user inputs must be validated server-side]
• [e.g. Pattern library: use <DataTable> component for all tabular data display]

These constraints must appear in the Architecture Constraints field of any story whose implementation path is affected. They will be checked by /review Category E and /definition-of-ready H9.

Any of these a blocker before decomposing? Reply: no - proceed / yes - [describe the issue]

If not found:

⚠️ No architecture-guardrails.md found - proceeding without guardrail check. Consider running /bootstrap or creating .github/architecture-guardrails.md from the template at .github/templates/architecture-guardrails.md to enable this check in future.

Record whether guardrails were available or absent in the epic artefact.

---

Step 2 - Choose a slicing strategy

Ask this before any decomposition. Do not default.

Which slicing strategy should I use for this feature?

1. Vertical slice — each story is a thin complete slice through all layers. Every story independently demo-able. Best for: high uncertainty, validating end-to-end behaviour early.

2. Walking skeleton — first story establishes the thinnest possible end-to-end path, subsequent stories flesh it out. Best for: new architectures or integrations needing proof before detail.

3. User journey — stories follow the user's chronological path through the feature. Best for: workflow-heavy features where sequence of interactions matters.

4. Risk-first — highest-risk or highest-uncertainty stories first. Best for: significant technical unknowns, de-risk before committing to full scope.

Reply: 1, 2, 3, or 4

Record the chosen strategy in every epic artefact — not implied, written explicitly.

---

Step 3 — Epic structure

Group stories into epics. Each epic is a cohesive body of work reviewable independently.

• Under ~8 stories: a single epic is fine
• Larger features: aim for 3–8 stories per epic

Present the proposed epic grouping before writing:

Proposed epic structure:

• Epic 1: [title] — [n stories] — [rationale for grouping]
• Epic 2: [title] — [n stories] — [rationale for grouping]

Does this grouping make sense, or do you want to reorganise? Reply: looks good — or describe how to reorganise

Save each epic to artefacts/[feature]/epics/[epic-slug].md conforming to .github/templates/epic.md.

---

Step 4 — Story decomposition

For each epic, write stories conforming to .github/templates/story.md.

Discipline:

• Every story names a persona from the benefit-metric artefact — not "a user"
• Every story's "So that..." connects to a named metric — not a feature preference
• Every story has a genuine out-of-scope section — not "N/A"
• Minimum 3 ACs per story in Given/When/Then format
• ACs describe observable behaviour, not implementation approach

Scope guard — per story: If a story is necessary but was not in the discovery MVP scope, surface it immediately rather than silently writing it:

⚠️ SCOPE NOTE: [Story title] was not in the discovery MVP scope.

It appears necessary because: [reason]

How do you want to handle it?

1. Add to MVP scope — I'll update the discovery artefact

Migration story detection: If a story is identified as a data migration, cutover, parallel-run, or consumer migration story (not user-facing, driven by data rules or traffic switching):

⚠️ This story type doesn't fit the standard story format. It looks like a [data migration / cutover / parallel-run / consumer migration] story.

Use migration-story.md template instead of story.md for this story?

1. Yes — use migration-story.md (appropriate for this type)
2. No — keep standard story.md format and I'll adapt the ACs

Reply: 1 or 2

If migration story template confirmed: write the story using .github/templates/migration-story.md.

2. Defer to post-MVP 3. Replace an existing MVP scope item — which one?

Reply: 1, 2, or 3

Save each story to artefacts/[feature]/stories/[story-slug].md

---

Step 5 — Benefit coverage matrix

After all stories are written, populate the metric coverage matrix in the benefit-metric artefact.

For each metric — list which stories move it.

Metric gap — surface immediately:

⚠️ METRIC GAP: [Metric name] has no stories that move it.

Options:

1. Write a story for it
2. Descope this metric
3. Mark as post-MVP

Reply: 1, 2, or 3

Story gap — surface immediately:

⚠️ STORY GAP: [Story title] has no metric linkage.

Either connect it to a metric or consider whether it belongs in MVP scope.

Which metric does this story move, or should it be removed? Reply: [metric name] — or remove

---

Step 6 — Scope accumulator

This step runs after all stories are written. It compares the total story scope against the original discovery MVP scope to detect drift that individual story guards may miss.

Individual stories can each be fine while the set together represents significant scope expansion. This step catches that.

Calculate:

• MVP scope item count from discovery artefact: [n]
• Total stories written: [n]
• Stories with scope notes (additions beyond MVP): [n]
• Stories explicitly deferred to post-MVP: [n]
• Stories with no metric linkage: [n]

Scope ratio check: For each discovery MVP scope item, confirm at least one story covers it. For each story, confirm it traces to at least one MVP scope item (or has an approved scope note).

Present the summary:

Scope accumulator — [feature name]

Discovery MVP scope items: [n] Stories written: [n] Coverage: [n of n MVP items covered] Scope additions (approved via scope note): [n] Scope ratio: [stories / MVP items] — [commentary]

[If ratio > 1.5 and additions not approved:] ⚠️ SCOPE DRIFT DETECTED The total story scope is [x]x the original discovery MVP. Scope additions: [list with their scope notes]

Is this intentional growth, or has scope crept?

1. Intentional — update the discovery artefact to reflect the expanded scope
2. Scope crept — I'll identify which stories to defer or remove
3. Correct — some stories cover multiple MVP items, ratio is misleading

Reply: 1, 2, or 3

[If all stories trace cleanly to MVP scope:]

✅ Scope check passed — [n] stories covering [n] MVP items. No unexplained scope additions detected.

---

Step 7 — NFR profile generation

After scope accumulator, generate the feature-level NFR profile.

Read from:

1. artefacts/[feature]/discovery.md — Constraints section
2. .github/product/constraints.md — hard product constraints (if exists)
3. Each story's NFR section — aggregate all story-level NFRs

Consolidate into a feature-level NFR profile:

NFR profile — [feature]:

Performance targets identified: [list or "None defined"] Security requirements: [list or "None defined"] Data classification: [Public / Internal / Confidential / Restricted] Data residency: [requirement or "Not applicable"] Availability SLA: [target or "Not defined"] Compliance frameworks: [list or "None"]

Any NFRs with named regulatory clauses require human sign-off before DoR.

Confirm this profile, or add/correct anything? Reply: confirm — or update [field] to [value]

Save to artefacts/[feature]/nfr-profile.md conforming to .github/templates/nfr-profile.md.

If no NFRs are identified, state this explicitly in the profile: Status: Active — No NFRs identified at definition. Reviewed at [date].

---

Quality checks before completing

• Every epic records its slicing strategy — written explicitly, not implied
• Every story's "So that..." connects to a named metric
• Every story has a genuine out-of-scope section (not "N/A")
• Minimum 3 ACs per story in Given/When/Then format
• Benefit coverage matrix complete — no orphaned metric, no unlinked story
• Scope accumulator run — ratio reviewed and any drift acknowledged
• All scope notes recorded in /decisions
• NFR profile generated and saved to artefacts/[feature]/nfr-profile.md

---

Completion output

Definition complete ✅

Epics: [n] at artefacts/[feature]/epics/ Stories: [n] at artefacts/[feature]/stories/ Slicing strategy: [chosen strategy] Scope check: ✅ Clean / ⚠️ [n] a

---

Content truncated.