Custom agent imported from B0yZ4kr14/OrthoPlus-Enterprise (.github/agents/speckit.sf.stories.agent.md). Copyright stays with the author.

Generate Developer Story Files

User Input

$ARGUMENTS

Supplemental Skill Discovery (Optional)

Before executing, you may check for any installed agent skills related to:

• Metadata/Story Analysis (sf-metadata, sf-data)
• Developer Workload Estimation (if specialized skills exist)

[!NOTE] These skills are Optional Accelerators. If found, use them to assist in decomposing features and identifying implementation layers. However, the standards in docs/scoring.md are the primary source of truth.

Prerequisites

• Plan exists: .specify/specs/NNN-feature-name/plan.md
• Data model exists: .specify/specs/NNN-feature-name/data-model.md
• Spec exists: .specify/specs/NNN-feature-name/spec.md
• Architect Sign-Off is completed in plan.md (check the sign-off section)
  • If sign-off is incomplete, WARN the user and ask if they want to proceed anyway

Instructions

Step 1: Read All Context

1. Read the plan from .specify/specs/NNN-feature-name/plan.md
2. Read the data model from .specify/specs/NNN-feature-name/data-model.md
3. Read the spec from .specify/specs/NNN-feature-name/spec.md
4. Read the constitution from .specify/memory/constitution.md

Step 2: Validate Architect Sign-Off

Check plan.md for the Architect Sign-Off section:

• If "Overall Sign-Off: [x] APPROVED" → proceed
• If not approved → warn: "⚠️ Architect has not signed off on this plan. Story generation may need revision after sign-off."
• Ask user whether to proceed or wait

Step 3: Generate Story-000 (Foundation Story)

ALWAYS generate this story first. It contains shared infrastructure that blocks all other stories.

Create .specify/specs/NNN-feature-name/task_story_00.md containing:

• Type: FULL
• What it covers:
  • All Custom Objects and Custom Fields referenced by ANY user story
  • Permission Sets with FLS for all new fields
  • Trigger files (one per object, with TAF MetadataTriggerHandler dispatch)
  • Test Data Factory class(es)
  • Custom Metadata for Trigger Action configuration
  • Base utility/selector classes shared across stories
• Dependencies: NONE (this is the root)
• Blocks: ALL other stories (task_story_01..NN)

Step 4: Decompose User Stories into Developer Stories

For each user story in the spec (US-1, US-2, US-3...):

1. Jira-Ready Description: Format as "As a [Persona], I want to [Action] so that [Business Benefit]".
2. Determine story type: FULL (has Apex/Flow/LWC) or DECLARATIVE (metadata-only)
3. Security & Access Matrix: Which Permission Sets/Profiles require access to which objects/fields.
4. Map to SF implementation layers: Apex classes, Flows, LWC components, test classes with exact file paths.
5. Detailed Acceptance Criteria: At least 3-5 Given/When/Then scenarios from the spec.
6. Generate test cases: Positive, Negative, and Bulk (251+ records).
7. Determine dependencies: Every story REQUIRES task_story_00.md.
8. Estimate effort: Hours for Human Developer Effort per layer.

Step 5: Create Story Files

For each story (01, 02, 03, ...):

Create .specify/specs/NNN-feature-name/task_story_NN.md:

1. Set Status to DRAFT
2. Set Story Type to FULL or DECLARATIVE
3. Populate SF Implementation Layers table with exact file paths
4. Populate Security & Access Matrix with detailed profile/permset configs
5. Populate Dependencies based on Step 4 analysis
6. Populate Scoring Gates with thresholds from the plan (mapped to the rubric in docs/scoring.md)
7. Populate Estimation table (Human Effort in hours)
8. Add TPO/Architect guidance in Developer Notes

Step 6: Build Dependency Graph

Create a summary showing the story dependency structure:

task_story_00.md (Foundation) ⛔ BLOCKS ALL
├── task_story_01.md (US-1: ...) [P]
├── task_story_02.md (US-2: ...) [P]
│   └── task_story_04.md (US-4: ...) — depends on 02
└── task_story_03.md (US-3: ...) [P]

[P] = Can work in parallel after Story-000 completes

Step 7: Present to User

Show the user:

• Number of stories generated (including Story-000)
• Dependency graph
• Which stories can be worked in parallel
• Total estimated effort (Human Hours)
• Recommendation for sprint planning

Next Step

Run /speckit.sf.review to have the Architect validate the story decomposition before creating Jira tickets.

Output

• Files created: task_story_00.md through task_story_NN.md
• Location: .specify/specs/NNN-feature-name/
• Status: All stories in DRAFT state

Notes

• Story-000 is ALWAYS generated and ALWAYS blocks other stories
• DECLARATIVE stories skip Apex/LWC layers and only include metadata scoring
• Mark parallel stories with [P] in the dependency graph
• Estimations represent Human Developer Effort in hours.