User Story Writing For Salesforce
This skill activates after requirements have been gathered and the BA must reshape raw business needs into INVEST-compliant Salesforce user stories that downstream admin/dev agents can pick up without re-asking the user. It owns the shape of a story (stem, acceptance criteria, sizing, splitting, handoff metadata) — not the elicitation upstream and not the UAT/build/prioritization downstream.
Before Starting
Gather this context before drafting stories:
- Has elicitation actually happened? A story without a real persona, real workflow, and real business outcome is a stub. If the requirement is still vague, return to
admin/requirements-gathering-for-sf before writing stories.
- Who is the persona, in Salesforce terms? "User" is not a persona. The persona must resolve to a profile, permission set, or role (e.g. Inside Sales Rep with Sales User profile). Stories without a Salesforce-grounded persona break sharing and FLS reasoning during build.
- Is there an existing story or epic this overlaps? Duplicate stories pollute the backlog, race for the same record, and waste sprint capacity. Search the existing backlog before drafting a new story.
- What downstream agents are available in the chain? The
recommended_agents[] handoff field only works if you know which agents the org runs (e.g. object-designer, flow-builder, lwc-builder, permission-set-architect). If unknown, default to ["object-designer", "permission-set-architect"] and flag in notes.
Core Concepts
The INVEST Checklist
Every story must pass INVEST before it leaves the BA's hands. Salesforce-specific reading:
- I — Independent. The story does not block on another incomplete story. Salesforce dependency hotspots: page layout depends on record type, validation rule depends on field, sharing rule depends on OWD. If the story has a hard dependency, list it in
dependencies[] and consider whether it can be split or reordered.
- N — Negotiable. Implementation detail (Flow vs Apex, screen vs record-triggered) is not fixed in the story body. The story states the what, the build agent picks the how. A story that names "use a Record-Triggered Flow with a fast field update" is over-specified.
- V — Valuable. The
So That must name a concrete business outcome (revenue, time saved, error reduced, compliance met). "So that the system works" or "so that data is captured" are not valuable.
- E — Estimable. The build team can size it. If complexity is unknowable because the persona, object, or trigger condition is missing, the story is not estimable and must go back to refinement.
- S — Small. Fits in a single sprint with capacity to spare. The complexity heuristic below makes this concrete.
- T — Testable. Acceptance criteria are boolean pass/fail. Each Given-When-Then must be runnable in a sandbox without interpretation.
The As-A / I-Want / So-That Stem
The canonical Salesforce shape:
As a [Salesforce-grounded persona — profile, permission set, or role],
I want [observable capability tied to a specific object/feature area],
So that [business outcome — revenue, time, error, compliance].
Anti-patterns to reject on sight:
- "As a user, I want…" — no persona.
- "As an admin, I want to add a field…" — describes the implementation, not the user need.
- "I want the system to automatically…" — system actions are not user wants; reshape as the persona who benefits.
- "So that the system works correctly" — not a business value statement.
Given-When-Then Acceptance Criteria
This skill uses Given-When-Then but does not teach it in depth (that's admin/acceptance-criteria-given-when-then). The minimum bar here:
- Every story has at least one Given-When-Then block.
- Each block is structured:
Given <context>, When <action>, Then <observable Salesforce outcome>.
- Observable outcome = a UI element, field value, error message, record state, or notification — not a database write or class invocation.
- Cover happy path + at least one sad path (validation failure, permission denial, null/empty case).
Complexity Sizing — S / M / L / XL
Use this heuristic when sizing — not story points, not Fibonacci, just a 4-bucket size that downstream agents can consume:
| Size |
Heuristic |
Typical example |
| S |
Single object, single field/perm change, no automation, no UI work. |
Add a required field; grant a permission set. |
| M |
Single object with one declarative automation (validation rule, simple flow), or one record page tweak. |
Auto-set a field via flow; restrict edit via VR. |
| L |
Multi-object, multi-step flow, or 1 LWC; may touch sharing or page layouts; single integration touchpoint. |
Lead-to-Opportunity conversion flow with related records. |
| XL |
Spans multiple personas, multi-system integration, or requires Apex/LWC + sharing redesign. An XL must be split before it leaves refinement. |
Quote-to-cash refactor; new approval routing across regions. |
XL stories are not committed — they are split. See the next section.
Story Splitting Techniques
When a story is XL (or arguably L but unclear), apply one of these splits in order of preference:
- By workflow steps. "Convert lead → create opportunity → notify rep" → 3 stories.
- By business rule. "Approval required for discount > 10% AND deal > $100k" → 2 stories, one rule per story.
- By data variation. "Works for Direct, Partner, and Channel deals" → 3 stories, one record type at a time.
- By persona. "Reps see X; managers see Y; execs see Z" → 3 stories per persona.
- By happy path / sad path. Ship the happy path first, then layer error/exception/edge stories.
Each child story still has to pass INVEST on its own. A split that produces children that can't be demoed independently is the wrong split.
Handoff Metadata — The JSON Shape
Every story emits a handoff JSON. Downstream agents (object-designer, flow-builder, lwc-builder, permission-set-architect) read this to pick up work without re-asking the user:
{
"story_id": "US-LEAD-042",
"title": "Auto-assign hot leads to inside sales queue",
"as_a": "Marketing Operations Specialist with the Marketing Ops profile",
"i_want": "incoming Leads with Score >= 80 to land in the Inside Sales queue automatically",
"so_that": "high-intent leads are followed up within the 5-minute SLA window",
"acceptance_criteria": [
"Given a Lead with Score >= 80, When it is created, Then OwnerId is set to the Inside Sales queue.",
"Given a Lead with Score < 80, When it is created, Then OwnerId remains the original creator.",
"Given a Lead update where Score crosses from 79 to 80, When saved, Then OwnerId reassigns to Inside Sales queue."
],
"complexity": "M",
"recommended_agents": ["flow-builder", "permission-set-architect"],
"recommended_skills": ["flow/record-triggered-flow-patterns", "admin/queue-design"],
"dependencies": ["US-LEAD-040 (Lead Score field exists)"],
"notes": "Confirm queue membership before flow build. No Apex required at this size."
}
Field rules:
recommended_agents[] — required, non-empty. Names the next agent(s) in the chain. If unsure, default to ["object-designer"] and flag in notes.
recommended_skills[] — pointers into the SfSkills repo (e.g. flow/record-triggered-flow-patterns). Optional but strongly preferred.
dependencies[] — story IDs or named prerequisites (field exists, queue exists, integration live).
complexity — exactly one of S | M | L | XL. XL means not committable — split first.
Common Patterns
Pattern: Reshape a Vague Requirement Into a Story
When to use: A stakeholder has said "we need lead routing." That's an epic, not a story.
How it works:
- Identify the persona — who actually performs (or benefits from) the work daily?
- Identify the trigger — what record event or user action starts the workflow?
- Identify the outcome — what observable state confirms it worked?
- Draft the As-A / I-Want / So-That stem.
- Add 3+ Given-When-Then criteria — happy + sad + edge.
- Size with the S/M/L/XL heuristic; split if XL.
- Emit the handoff JSON with
recommended_agents[] populated.
Why not skip steps: A story with no Given-When-Then is a wish, not a backlog item. Build agents will refuse to pick it up.
Pattern: Split an XL Story
When to use: Sizing came back XL, or the team can't agree on size (which usually means it's XL with hidden scope).
How it works:
- Pick the split axis (workflow, rule, data, persona, happy/sad).
- Draft each child story with its own stem + Given-When-Then.
- Verify each child passes INVEST independently.
- Order the children: happy path first, sad path last; high-value first when possible.
- Add cross-references in
dependencies[].
- Replace the parent story with a tracker/epic; do not commit it.
Why not just commit the XL: XL stories drag across sprints, block dependent work, and demo poorly. The split is the correctness step.
Pattern: Emit Handoff JSON Alongside Markdown
When to use: Always. The markdown is for humans; the JSON is for the next agent.
How it works:
- Write the story as markdown for backlog tools (Jira, ADO, GitHub Issues).
- Below the markdown, emit the JSON block in a fenced code block.
- Populate
recommended_agents[] from the chain available in the org.
- Validate the JSON with
scripts/check_invest.py (this skill's checker) before finalizing.
Decision Guidance
| Situation |
Recommended Approach |
Reason |
| Persona is "user" or "admin" |
Reject and ask BA to ground in a Salesforce profile/perm set/role |
Generic personas break sharing and FLS reasoning |
| Story body names Flow vs Apex |
Reject — that's a build-time decision |
Violates INVEST-Negotiable |
| Acceptance criteria test the UI ("the button is blue") |
Rewrite to test behavior ("clicking Save creates the record") |
UI is Salesforce's responsibility; ACs test outcomes |
| Story has only happy-path AC |
Add at least one sad path before sizing |
Half-tested stories regress in UAT |
| Story sizes to XL |
Split before committing |
XL is uncommittable by definition |
| Story depends on another in-flight story |
Add to dependencies[]; consider reordering |
Independent-I in INVEST |
| Story mixes "the user does X" and "the system does Y" |
Split by action source |
Mixing actors hides automation triggers |
recommended_agents[] is unknown |
Default to ["object-designer"] and flag in notes |
Empty handoff metadata breaks the chain |
Recommended Workflow
Step-by-step instructions an AI agent or BA follows when this skill activates:
- Confirm prerequisites — verify requirements were gathered (skill:
admin/requirements-gathering-for-sf). If not, route there first.
- Draft the stem — As-A (Salesforce-grounded persona), I-Want (observable capability), So-That (business outcome). Reject generic personas and "the system" wants.
- Author Given-When-Then ACs — 1 happy path minimum, 1+ sad path, edge cases as applicable. Each AC must be boolean-testable in a sandbox.
- Size with S/M/L/XL — apply the heuristic table. XL means not committable.
- Split if needed — pick a split axis, draft child stories, verify each child passes INVEST independently.
- Emit handoff JSON — populate
recommended_agents[], recommended_skills[], dependencies[], complexity, notes. Validate with scripts/check_invest.py.
- Run review checklist — INVEST pass, persona grounded, AC count > 0, complexity present, handoff JSON valid.
Review Checklist
Salesforce-Specific Gotchas
- Page-layout and record-type chains — A story to "show a new field on the layout" is not Independent if the field doesn't exist yet. The build order is field → layout → validation rule. Surface this in
dependencies[] or split the parent.
- Sharing-rule stories that look small — "Managers can see their team's Cases" sounds S, but if OWD is Public Read/Write, the answer is "they already can" and the real story is OWD tightening. Sizing without checking OWD produces wrong-sized stories.
- Persona drift between stem and AC — Stem says "Sales Rep" but the AC tests "Sales Manager can approve." That's two stories; split by persona.
Output Artifacts
| Artifact |
Description |
| Story markdown |
Human-readable As-A / I-Want / So-That + Given-When-Then ACs, sized S/M/L/XL |
| Handoff JSON |
Machine-readable shape consumed by object-designer, flow-builder, lwc-builder, permission-set-architect |
| Split rationale |
When an XL is split into children, a short note explaining the split axis chosen |
| INVEST self-check report |
Output of scripts/check_invest.py confirming the story passes structural lints |
Related Skills
admin/requirements-gathering-for-sf — runs upstream; produces the raw needs this skill reshapes
admin/acceptance-criteria-given-when-then — owns the GWT technique in depth; this skill only enforces presence
admin/uat-test-case-design — runs downstream; turns ACs into UAT scripts after build
admin/moscow-prioritization-for-sf-backlog — runs alongside; orders the stories this skill produces
agents/object-designer/AGENT.md, agents/flow-builder/AGENT.md, agents/lwc-builder/AGENT.md, agents/permission-set-architect/AGENT.md — common downstream consumers of the handoff JSON
1---2name: user-story-writing-for-salesforce3description: "Use this skill when authoring INVEST-compliant Salesforce user stories from gathered requirements: shaping the As-A/I-Want/So-That stem, writing Given-When-Then acceptance criteria, sizing complexity (S/M/L/XL), splitting stories that are too large for a sprint, and emitting handoff metadata that names the next downstream agent. Trigger keywords: user story, INVEST, story splitting, story sizing, As-A I-Want So-That, story handoff, recommended_agents, backlog item shape. NOT for requirements elicitation or stakeholder interviews (use admin/requirements-gathering-for-sf). NOT for the Given-When-Then technique itself in depth (use admin/acceptance-criteria-given-when-then). NOT for UAT test case design (use admin/uat-test-case-design). NOT for backlog prioritization (use admin/moscow-prioritization-for-sf-backlog)."4category: admin5salesforce-version: "Spring '25+"6well-architected-pillars:7 - Operational Excellence8 - User Experience9 - Reliability10triggers:11 - "how do I write a Salesforce user story that an admin or developer can actually pick up"12 - "my user stories are too big to finish in a sprint, how do I split them"13 - "what is the canonical As-A I-Want So-That format for a Salesforce backlog item"14 - "how do I add Given-When-Then acceptance criteria to a story"15 - "what handoff metadata should a story carry so the right downstream agent picks it up"16 - "how do I size a Salesforce user story when complexity is unclear"17 - "how do I check that my user stories pass the INVEST test"18 - "user stories isn't working"19tags:20 - user-stories21 - invest22 - story-splitting23 - story-sizing24 - acceptance-criteria25 - agile-handoff26 - backlog-shape27inputs:28 - "Gathered requirement, feature description, or epic narrative to be reshaped into stories"29 - "Target persona(s) — Salesforce profile, permission set, or role of the end user"30 - "Known platform constraints (object volume, integration deps, sharing model) that affect splitting"31 - "Available downstream agents in the chain (e.g. object-designer, flow-builder, lwc-builder, permission-set-architect)"32outputs:33 - "Set of INVEST-compliant Salesforce user stories in canonical As-A/I-Want/So-That + Given-When-Then markdown"34 - "Handoff JSON for each story with story_id, title, as_a, i_want, so_that, acceptance_criteria[], complexity, recommended_agents[], recommended_skills[], dependencies[], notes"35 - "Story-split rationale when an oversize story is broken into 2+ children"36 - "Complexity sizing per story (S / M / L / XL) with the heuristic that drove the score"37dependencies: []38version: 1.0.039author: Pranav Nagrecha40updated: 2026-04-2841metadata:42 source: https://skillmd.com/skills/pranavnagrecha/user-story-writing-for-salesforce43---4445# User Story Writing For Salesforce4647This skill activates after requirements have been gathered and the BA must reshape raw business needs into INVEST-compliant Salesforce user stories that downstream admin/dev agents can pick up without re-asking the user. It owns the *shape* of a story (stem, acceptance criteria, sizing, splitting, handoff metadata) — not the elicitation upstream and not the UAT/build/prioritization downstream.4849---5051## Before Starting5253Gather this context before drafting stories:5455- **Has elicitation actually happened?** A story without a real persona, real workflow, and real business outcome is a stub. If the requirement is still vague, return to `admin/requirements-gathering-for-sf` before writing stories.56- **Who is the persona, in Salesforce terms?** "User" is not a persona. The persona must resolve to a profile, permission set, or role (e.g. *Inside Sales Rep with Sales User profile*). Stories without a Salesforce-grounded persona break sharing and FLS reasoning during build.57- **Is there an existing story or epic this overlaps?** Duplicate stories pollute the backlog, race for the same record, and waste sprint capacity. Search the existing backlog before drafting a new story.58- **What downstream agents are available in the chain?** The `recommended_agents[]` handoff field only works if you know which agents the org runs (e.g. `object-designer`, `flow-builder`, `lwc-builder`, `permission-set-architect`). If unknown, default to `["object-designer", "permission-set-architect"]` and flag in `notes`.5960---6162## Core Concepts6364### The INVEST Checklist6566Every story must pass INVEST before it leaves the BA's hands. Salesforce-specific reading:6768- **I — Independent.** The story does not block on another incomplete story. Salesforce dependency hotspots: page layout depends on record type, validation rule depends on field, sharing rule depends on OWD. If the story has a hard dependency, list it in `dependencies[]` and consider whether it can be split or reordered.69- **N — Negotiable.** Implementation detail (Flow vs Apex, screen vs record-triggered) is *not* fixed in the story body. The story states the *what*, the build agent picks the *how*. A story that names "use a Record-Triggered Flow with a fast field update" is over-specified.70- **V — Valuable.** The `So That` must name a concrete business outcome (revenue, time saved, error reduced, compliance met). "So that the system works" or "so that data is captured" are not valuable.71- **E — Estimable.** The build team can size it. If complexity is unknowable because the persona, object, or trigger condition is missing, the story is not estimable and must go back to refinement.72- **S — Small.** Fits in a single sprint with capacity to spare. The complexity heuristic below makes this concrete.73- **T — Testable.** Acceptance criteria are boolean pass/fail. Each Given-When-Then must be runnable in a sandbox without interpretation.7475### The As-A / I-Want / So-That Stem7677The canonical Salesforce shape:7879```80As a [Salesforce-grounded persona — profile, permission set, or role],81I want [observable capability tied to a specific object/feature area],82So that [business outcome — revenue, time, error, compliance].83```8485Anti-patterns to reject on sight:86- "As a user, I want…" — no persona.87- "As an admin, I want to add a field…" — describes the *implementation*, not the user need.88- "I want the system to automatically…" — system actions are not user wants; reshape as the persona who benefits.89- "So that the system works correctly" — not a business value statement.9091### Given-When-Then Acceptance Criteria9293This skill uses Given-When-Then but does **not** teach it in depth (that's `admin/acceptance-criteria-given-when-then`). The minimum bar here:9495- Every story has **at least one** Given-When-Then block.96- Each block is structured: `Given <context>`, `When <action>`, `Then <observable Salesforce outcome>`.97- Observable outcome = a UI element, field value, error message, record state, or notification — not a database write or class invocation.98- Cover **happy path + at least one sad path** (validation failure, permission denial, null/empty case).99100### Complexity Sizing — S / M / L / XL101102Use this heuristic when sizing — not story points, not Fibonacci, just a 4-bucket size that downstream agents can consume:103104| Size | Heuristic | Typical example |105|---|---|---|106| **S** | Single object, single field/perm change, no automation, no UI work. | Add a required field; grant a permission set. |107| **M** | Single object with one declarative automation (validation rule, simple flow), or one record page tweak. | Auto-set a field via flow; restrict edit via VR. |108| **L** | Multi-object, multi-step flow, or 1 LWC; may touch sharing or page layouts; single integration touchpoint. | Lead-to-Opportunity conversion flow with related records. |109| **XL** | Spans multiple personas, multi-system integration, or requires Apex/LWC + sharing redesign. **An XL must be split** before it leaves refinement. | Quote-to-cash refactor; new approval routing across regions. |110111XL stories are not committed — they are split. See the next section.112113### Story Splitting Techniques114115When a story is XL (or arguably L but unclear), apply one of these splits in order of preference:1161171. **By workflow steps.** "Convert lead → create opportunity → notify rep" → 3 stories.1182. **By business rule.** "Approval required for discount > 10% AND deal > $100k" → 2 stories, one rule per story.1193. **By data variation.** "Works for Direct, Partner, and Channel deals" → 3 stories, one record type at a time.1204. **By persona.** "Reps see X; managers see Y; execs see Z" → 3 stories per persona.1215. **By happy path / sad path.** Ship the happy path first, then layer error/exception/edge stories.122123Each child story still has to pass INVEST on its own. A split that produces children that can't be demoed independently is the wrong split.124125### Handoff Metadata — The JSON Shape126127Every story emits a handoff JSON. Downstream agents (`object-designer`, `flow-builder`, `lwc-builder`, `permission-set-architect`) read this to pick up work without re-asking the user:128129```json130{131 "story_id": "US-LEAD-042",132 "title": "Auto-assign hot leads to inside sales queue",133 "as_a": "Marketing Operations Specialist with the Marketing Ops profile",134 "i_want": "incoming Leads with Score >= 80 to land in the Inside Sales queue automatically",135 "so_that": "high-intent leads are followed up within the 5-minute SLA window",136 "acceptance_criteria": [137 "Given a Lead with Score >= 80, When it is created, Then OwnerId is set to the Inside Sales queue.",138 "Given a Lead with Score < 80, When it is created, Then OwnerId remains the original creator.",139 "Given a Lead update where Score crosses from 79 to 80, When saved, Then OwnerId reassigns to Inside Sales queue."140 ],141 "complexity": "M",142 "recommended_agents": ["flow-builder", "permission-set-architect"],143 "recommended_skills": ["flow/record-triggered-flow-patterns", "admin/queue-design"],144 "dependencies": ["US-LEAD-040 (Lead Score field exists)"],145 "notes": "Confirm queue membership before flow build. No Apex required at this size."146}147```148149Field rules:150- `recommended_agents[]` — **required, non-empty**. Names the next agent(s) in the chain. If unsure, default to `["object-designer"]` and flag in `notes`.151- `recommended_skills[]` — pointers into the SfSkills repo (e.g. `flow/record-triggered-flow-patterns`). Optional but strongly preferred.152- `dependencies[]` — story IDs or named prerequisites (field exists, queue exists, integration live).153- `complexity` — exactly one of `S | M | L | XL`. XL means *not committable* — split first.154155---156157## Common Patterns158159### Pattern: Reshape a Vague Requirement Into a Story160161**When to use:** A stakeholder has said "we need lead routing." That's an epic, not a story.162163**How it works:**1641. Identify the persona — who actually performs (or benefits from) the work daily?1652. Identify the trigger — what record event or user action starts the workflow?1663. Identify the outcome — what observable state confirms it worked?1674. Draft the As-A / I-Want / So-That stem.1685. Add 3+ Given-When-Then criteria — happy + sad + edge.1696. Size with the S/M/L/XL heuristic; split if XL.1707. Emit the handoff JSON with `recommended_agents[]` populated.171172**Why not skip steps:** A story with no Given-When-Then is a wish, not a backlog item. Build agents will refuse to pick it up.173174### Pattern: Split an XL Story175176**When to use:** Sizing came back XL, or the team can't agree on size (which usually means it's XL with hidden scope).177178**How it works:**1791. Pick the split axis (workflow, rule, data, persona, happy/sad).1802. Draft each child story with its own stem + Given-When-Then.1813. Verify each child passes INVEST independently.1824. Order the children: happy path first, sad path last; high-value first when possible.1835. Add cross-references in `dependencies[]`.1846. Replace the parent story with a tracker/epic; do not commit it.185186**Why not just commit the XL:** XL stories drag across sprints, block dependent work, and demo poorly. The split is the correctness step.187188### Pattern: Emit Handoff JSON Alongside Markdown189190**When to use:** Always. The markdown is for humans; the JSON is for the next agent.191192**How it works:**1931. Write the story as markdown for backlog tools (Jira, ADO, GitHub Issues).1942. Below the markdown, emit the JSON block in a fenced code block.1953. Populate `recommended_agents[]` from the chain available in the org.1964. Validate the JSON with `scripts/check_invest.py` (this skill's checker) before finalizing.197198---199200## Decision Guidance201202| Situation | Recommended Approach | Reason |203|---|---|---|204| Persona is "user" or "admin" | Reject and ask BA to ground in a Salesforce profile/perm set/role | Generic personas break sharing and FLS reasoning |205| Story body names Flow vs Apex | Reject — that's a build-time decision | Violates INVEST-Negotiable |206| Acceptance criteria test the UI ("the button is blue") | Rewrite to test behavior ("clicking Save creates the record") | UI is Salesforce's responsibility; ACs test outcomes |207| Story has only happy-path AC | Add at least one sad path before sizing | Half-tested stories regress in UAT |208| Story sizes to XL | Split before committing | XL is uncommittable by definition |209| Story depends on another in-flight story | Add to `dependencies[]`; consider reordering | Independent-I in INVEST |210| Story mixes "the user does X" and "the system does Y" | Split by action source | Mixing actors hides automation triggers |211| `recommended_agents[]` is unknown | Default to `["object-designer"]` and flag in `notes` | Empty handoff metadata breaks the chain |212213---214215## Recommended Workflow216217Step-by-step instructions an AI agent or BA follows when this skill activates:2182191. **Confirm prerequisites** — verify requirements were gathered (skill: `admin/requirements-gathering-for-sf`). If not, route there first.2202. **Draft the stem** — As-A (Salesforce-grounded persona), I-Want (observable capability), So-That (business outcome). Reject generic personas and "the system" wants.2213. **Author Given-When-Then ACs** — 1 happy path minimum, 1+ sad path, edge cases as applicable. Each AC must be boolean-testable in a sandbox.2224. **Size with S/M/L/XL** — apply the heuristic table. XL means not committable.2235. **Split if needed** — pick a split axis, draft child stories, verify each child passes INVEST independently.2246. **Emit handoff JSON** — populate `recommended_agents[]`, `recommended_skills[]`, `dependencies[]`, `complexity`, `notes`. Validate with `scripts/check_invest.py`.2257. **Run review checklist** — INVEST pass, persona grounded, AC count > 0, complexity present, handoff JSON valid.226227---228229## Review Checklist230231- [ ] Every story has a Salesforce-grounded persona (profile / permission set / role) — not "user" or "admin"232- [ ] Every `So That` names a concrete business outcome (revenue, time, error, compliance)233- [ ] Every story has at least one Given-When-Then acceptance criterion234- [ ] At least one sad path AC is present per story235- [ ] No story body names a specific Flow type, Apex class, or LWC component (Negotiable)236- [ ] Complexity is exactly one of S / M / L / XL237- [ ] No story is committed at XL — XL is split first238- [ ] Handoff JSON is present, valid, and `recommended_agents[]` is non-empty239- [ ] Dependencies on other stories are explicit in `dependencies[]`240- [ ] Story word count is reasonable (under ~250 words for the body); if it isn't, split241242---243244## Salesforce-Specific Gotchas2452461. **Page-layout and record-type chains** — A story to "show a new field on the layout" is not Independent if the field doesn't exist yet. The build order is field → layout → validation rule. Surface this in `dependencies[]` or split the parent.2472. **Sharing-rule stories that look small** — "Managers can see their team's Cases" sounds S, but if OWD is Public Read/Write, the answer is "they already can" and the real story is OWD tightening. Sizing without checking OWD produces wrong-sized stories.2483. **Persona drift between stem and AC** — Stem says "Sales Rep" but the AC tests "Sales Manager can approve." That's two stories; split by persona.249250---251252## Output Artifacts253254| Artifact | Description |255|---|---|256| Story markdown | Human-readable As-A / I-Want / So-That + Given-When-Then ACs, sized S/M/L/XL |257| Handoff JSON | Machine-readable shape consumed by `object-designer`, `flow-builder`, `lwc-builder`, `permission-set-architect` |258| Split rationale | When an XL is split into children, a short note explaining the split axis chosen |259| INVEST self-check report | Output of `scripts/check_invest.py` confirming the story passes structural lints |260261---262263## Related Skills264265- `admin/requirements-gathering-for-sf` — runs upstream; produces the raw needs this skill reshapes266- `admin/acceptance-criteria-given-when-then` — owns the GWT technique in depth; this skill only enforces presence267- `admin/uat-test-case-design` — runs downstream; turns ACs into UAT scripts after build268- `admin/moscow-prioritization-for-sf-backlog` — runs alongside; orders the stories this skill produces269- `agents/object-designer/AGENT.md`, `agents/flow-builder/AGENT.md`, `agents/lwc-builder/AGENT.md`, `agents/permission-set-architect/AGENT.md` — common downstream consumers of the handoff JSON