Skip to main content

/jaan-to:dev-be-task-breakdown

Convert PRDs into structured backend development tasks with data model notes, reliability patterns, and error taxonomy.

What It Does​

Transforms a PRD into a comprehensive backend task breakdown with:

  • Entity extraction - Nouns → models, verbs → controllers, temporal → jobs
  • Implicit task detection - DB indexes, error handling, security, observability
  • Master task cards - Each task includes: data model notes, idempotency patterns, error scenarios, reliability notes
  • Dependency mapping - Critical path, parallel tracks, blocked-by chains
  • Framework adaptation - Reads tech.md to adapt for Laravel, Django, FastAPI, etc.
  • 3 export formats - Jira CSV, Linear markdown, JSON

Uses vertical slicing by default (user-facing features, not technical layers).

Usage​

/jaan-to:dev-be-task-breakdown "path/to/prd.md"

Or with a feature description:

/jaan-to:dev-be-task-breakdown "User can subscribe to premium features"

Or with a tech plan (from upstream skill):

/jaan-to:dev-be-task-breakdown "path/to/tech-plan.md"

What It Asks​

QuestionWhy
Slicing strategyVertical (features) or horizontal (layers)?
Team sizeCalibrates T-shirt sizes (2-dev vs 8-dev team)
API conventionsREST/JSON, GraphQL, gRPC?
Delete strategyHard delete or soft delete?
Legacy constraintsMixing old and new tech stacks?

Then shows a HARD STOP with:

  • Entity map (models, relationships, task count)
  • Task list preview (ID, type, size, dependencies)
  • Implicit tasks detected
  • Critical path
  • XXL task warnings (if any need decomposition)

Output​

Path: $JAAN_OUTPUTS_DIR/dev/backend/{slug}/task-breakdown.md

Structure:

  • Overview table (metrics)
  • Tech stack (imported from tech.md)
  • Entity map
  • Task cards (full breakdown with 8+ sections each)
  • Dependency graph
  • Critical path analysis
  • Implicit tasks (auto-detected)
  • Ambiguity defaults applied
  • 3 export formats (Jira CSV, Linear MD, JSON)
  • Validation summary
  • Definition of Ready/Done

Task Card Template: Each task includes:

  1. ID, Type, Size, Priority, Files, Dependencies
  2. Acceptance Criteria (3-5 testable bullets)
  3. Data Model Notes (table, columns, indexes, constraints)
  4. Idempotency (type, key, storage, duplicate handling)
  5. Error Scenarios (scenario → HTTP code → handling)
  6. Reliability Notes (queue config, retries, circuit breaker)
  7. Security Checklist
  8. Test Requirements

Example​

Input:

/jaan-to:dev-be-task-breakdown "$JAAN_OUTPUTS_DIR/pm/user-subscriptions/prd.md"

Extracts:

  • Entities: User, Subscription, Plan, Payment
  • Actions: create subscription, cancel subscription, upgrade plan
  • Jobs: Process payment, send welcome email, sync to billing system
  • Policies: User can only cancel own subscription

Generates:

  • 12 vertical-slice tasks
  • 4 implicit tasks (indexes, error handling, audit logs, monitoring)
  • Critical path: 7 tasks (sequential)
  • Parallel tracks: 5 tasks (can run concurrently)
  • T-shirt sizes: 2 XS, 6 S, 3 M, 1 L
  • Estimated duration: 3-4 sprints (2-week sprints, 2-dev team)

Output: $JAAN_OUTPUTS_DIR/dev/backend/user-subscriptions/task-breakdown.md

Tips​

  • Read tech.md first - Framework determines task templates (Laravel ≠ Django)
  • Vertical slicing wins - "Add subscription" (migration + model + controller + tests) beats "Create all migrations"
  • Implicit tasks matter - PRDs forget indexes, error handling, security
  • HARD STOP is critical - Catching entity extraction mistakes early saves hours
  • Use tech plan upstream - /jaan-to:dev-tech-plan → this skill = better task breakdown
  • Export all 3 formats - Costs nothing, saves time when switching project trackers
  • Watch for XXL warnings - Tasks >800 LOC should be decomposed