Spec vers tâches avec MCP : convertir une spec fonctionnelle en arbre de tâches
Transform feature specifications into structured task trees using MCP. This guide shows you how to parse a pasted spec, create a task outline, add estimates and acceptance criteria, and create tasks using a two-pass approach with confirmation.
Ce que ce workflow permet
Spec to tasks workflow with MCP converts specifications into actionable work:
Résultats clés
- Spec parsing: Analyze feature specifications and extract requirements
- Task breakdown: Convert specs into structured task trees
- Estimation: Add time estimates to tasks
- Acceptance criteria: Generate acceptance criteria for each task
- Two-pass creation: Outline first, then create tasks after confirmation
Prérequis
Before using this workflow, ensure you have:
- Corcava MCP server configured in your AI assistant
- API key with write access to tasks
- Feature specification ready to paste
- Project ID where tasks will be created
Workflow pas à pas
Step 1: Provide Feature Spec
Start by pasting your feature specification:
Spec Input Prompt
What the AI does:
- Analyzes the feature specification
- Identifies components and requirements
- Breaks down into logical tasks
- Creates task hierarchy (parent/child relationships)
- Presents task outline for review
Step 2: Review Task Outline
Review and refine the proposed task structure:
Outline Refinement Prompt
What the AI does:
- Modifies task outline based on feedback
- Shows updated structure
- Waits for approval before creating tasks
Step 3: Add Estimates and Acceptance Criteria
Enhance tasks with estimates and criteria:
Enhancement Prompt
What the AI does:
- Analyzes each task for complexity
- Estimates time required
- Generates acceptance criteria
- Identifies task dependencies
- Presents enhanced task list
Step 4: Create Tasks (Two-Pass Approach with Confirmation)
⚠️ Always Confirm Before Creating Tasks
When creating tasks from specs:
- Outline First: Show complete task structure before creating
- Confirmation Token: Require "CONFIRM ALL" for batch creation
- Review Details: Display title, description, estimate, acceptance criteria for each task
- Verify Project: Confirm tasks go to the correct project
Safe Task Creation Prompt
What the AI does:
- Shows complete task list with all details
- Confirms project name and ID
- Waits for "CONFIRM ALL"
- Only after confirmation, calls
create_taskfor each task - Confirms each task was created successfully
Step 5: Add Acceptance Criteria as Comments
Post acceptance criteria as comments on created tasks:
Criteria Posting Prompt
What the AI does:
- Shows preview of all acceptance criteria comments
- Waits for "CONFIRM"
- Only after confirmation, calls
add_task_commentfor each task
Complete Workflow Prompts
Here are complete, copy-paste ready prompts:
Full Spec to Tasks Workflow
Use this when: You want a complete two-pass workflow
Quick Spec Breakdown
Use this when: You want to review the breakdown before creating
Workflow Variants
Variant 1: Hierarchical Task Tree
Create parent-child task relationships:
What happens: AI creates structured task hierarchy
Variant 2: Sequential Task Flow
Order tasks by execution sequence:
What happens: AI creates sequential task flow with dependencies
Example: Complete Spec to Tasks Session
Here's a real example of how a spec to tasks session flows:
User Prompt
AI Response (Task Outline)
Proposed Task Breakdown (6 tasks):
- Task 1: Design user dashboard layout
Estimate: 4 hours | Acceptance: Layout mockups approved - Task 2: Implement user profile display component
Estimate: 6 hours | Acceptance: Profile displays correctly with all fields - Task 3: Build activity feed component
Estimate: 8 hours | Acceptance: Feed shows last 10 activities, updates in real-time - Task 4: Add quick action buttons
Estimate: 3 hours | Acceptance: Buttons trigger correct actions - Task 5: Implement responsive mobile layout
Estimate: 6 hours | Acceptance: Dashboard works on mobile devices - Task 6: Write tests for dashboard components
Estimate: 4 hours | Acceptance: All components have unit tests
Ready to create? Type CONFIRM ALL to create all 6 tasks in project 'User Dashboard'.
Safety Patterns
⚠️ Confirmation Rules for Task Creation
When creating tasks from specs:
- Always outline first: Show complete task structure before creating
- Require explicit confirmation: Use "CONFIRM ALL" for batch creation
- Review details: Verify estimates, acceptance criteria, and project assignment
- Two-pass approach: Outline → Review → Enhance → Confirm → Create
Troubleshooting
Tasks Too Granular
Symptom: AI created too many small tasks
Possible causes:
- Spec was too detailed
- AI broke down too much
Fix: Ask for consolidation: "Merge related tasks, keep it to 5-7 main tasks"
Missing Requirements
Symptom: Some spec requirements not in tasks
Possible causes:
- Spec was ambiguous
- AI missed implicit requirements
Fix: Provide clarification: "Also include tasks for [missing requirement]"
Related Tools
This workflow uses these Corcava MCP tools:
Related Use Cases
Sprint Planning
Plan sprints from spec-derived tasks
Acceptance Criteria
Add acceptance criteria to spec tasks
Convert Specs to Tasks Automatically
Connect your AI assistant to Corcava and transform feature specs into actionable tasks
