Corcava logoLe seul outil métier dont vous avez besoinCorcava
Menu
Solutions
Secteurs
DesignersRH & recrutementCabinets d'avocatsImmobilierConseilMarketing digitalComptabilitéE‑commerceDéveloppement logiciel
Produit
FonctionnalitésIntégrations
BlogTélécharger
Ressources
TemplatesDocsMCPFAQAPIComparaisonsFeuille de routeQuoi de neuf$ Programme d'affiliation
Tarifs
Essai gratuit de 14 jours

Aucune carte bancaire requise

Critères d'acceptation avec MCP : transformer les tâches vagues en résultats testables

Transform vague tasks into testable acceptance criteria using MCP. This playbook shows you how to read a task, propose clear acceptance criteria, and add them as a comment with confirmation—with examples for engineering, design, and ops tasks.

Ce que ce workflow permet

Acceptance criteria workflow with MCP clarifies task requirements:

Résultats clés

  • Task analysis: Read and understand task context and requirements
  • Criteria generation: Propose clear, testable acceptance criteria
  • Domain-specific examples: Tailor criteria for engineering, design, and ops tasks
  • Documentation: Add acceptance criteria as comments for tracking
  • Quality improvement: Transform vague tasks into actionable, testable work

Prérequis

Before using this workflow, ensure you have:

Workflow pas à pas

Step 1: Read the Task

Start by getting the full task context:

Task Reading Prompt

"Read task [Task ID or Title] and analyze what needs to be done. Show me the task title, description, current status, and any existing comments."

What the AI does:

  1. Calls get_task to retrieve task details
  2. Calls list_task_comments to get existing context
  3. Analyzes task description and requirements
  4. Presents task context for analysis

Step 2: Analyze Requirements

Identify what needs to be clarified:

Requirements Analysis Prompt

"For this task, identify: - What is the expected outcome? - What are the unclear or vague requirements? - What would make this task 'done'? - What edge cases or scenarios should be considered? Show me your analysis before proposing acceptance criteria."

What the AI does:

  1. Analyzes task description for clarity
  2. Identifies vague or ambiguous requirements
  3. Determines expected outcomes
  4. Considers edge cases and scenarios

Step 3: Propose Acceptance Criteria

Generate testable acceptance criteria:

Criteria Generation Prompt

"Based on the task analysis, propose 3-5 clear, testable acceptance criteria. Each criterion should: - Be specific and measurable - Define what 'done' looks like - Be testable (can be verified) - Cover the main requirements and edge cases Format them as a numbered list. Show me the criteria before adding them as a comment."

What the AI does:

  1. Generates specific, testable criteria
  2. Ensures criteria are measurable
  3. Covers main requirements and edge cases
  4. Formats as clear list
  5. Presents criteria for review

Step 4: Add Criteria as Comment (with Confirmation)

⚠️ Always Confirm Before Posting Comments

When posting acceptance criteria as comments:

  • Preview First: Show comment text before posting
  • Confirmation Token: Require "CONFIRM" before adding comment
  • Verify Task: Confirm which task the comment will be added to

Safe Comment Posting Prompt

"Add the acceptance criteria as a comment on task [Task ID]. But first, show me exactly what comment you'll post. Format it clearly with a heading 'Acceptance Criteria:' followed by the numbered list. Wait for CONFIRM before adding the comment."

What the AI does:

  1. Formats acceptance criteria as comment
  2. Shows comment preview
  3. Waits for "CONFIRM"
  4. Only after confirmation, calls add_task_comment
  5. Confirms comment was posted

Complete Workflow Prompts

Here are complete, copy-paste ready prompts:

Full Acceptance Criteria Workflow

"For task [Task ID]: 1. Read the task and analyze requirements 2. Identify what's vague or unclear 3. Propose 3-5 testable acceptance criteria 4. Show me the criteria 5. After I approve, format them as a comment 6. Show me the comment you'll post 7. Wait for CONFIRM 8. Add the comment to the task Important: Don't add the comment until I type CONFIRM."

Use this when: You want a complete workflow with full control

Quick Criteria Generation

"Read task [Task ID] and propose 3-5 testable acceptance criteria. Show me the criteria but don't add them as a comment yet."

Use this when: You want to review criteria before posting

Workflow Variants by Domain

Variant 1: Engineering Tasks

Focus on technical, testable criteria:

"For this engineering task, propose acceptance criteria that are: - Technically specific (APIs, data structures, behaviors) - Testable with unit/integration tests - Include edge cases and error handling - Define performance requirements if applicable Focus on what can be verified through code or tests."

What happens: AI generates technical, code-testable criteria

Variant 2: Design Tasks

Focus on user experience and visual criteria:

"For this design task, propose acceptance criteria that cover: - User experience requirements - Visual design specifications - Responsive behavior (mobile, tablet, desktop) - Accessibility requirements - Design system compliance Focus on what can be verified through design review or user testing."

What happens: AI generates UX/design-focused criteria

Variant 3: Operations Tasks

Focus on operational and process criteria:

"For this ops task, propose acceptance criteria that cover: - Process completion requirements - Documentation needs - Monitoring and alerting setup - Rollback procedures - Success metrics Focus on what can be verified through process checks or metrics."

What happens: AI generates operational, process-focused criteria

Example: Complete Acceptance Criteria Session

Here's a real example of how an acceptance criteria session flows:

User Prompt

"For task 'Implement user authentication', propose acceptance criteria. Show me the criteria first, then wait for CONFIRM before adding as a comment."

AI Response (Criteria Proposal)

Proposed Acceptance Criteria:

  1. User can register with email and password (minimum 8 characters, must include uppercase, lowercase, and number)
  2. User can log in with valid credentials and receive an authentication token
  3. User cannot log in with invalid credentials (wrong password, non-existent email)
  4. Authentication token expires after 24 hours and user must re-authenticate
  5. User can reset password via email link
  6. All authentication endpoints return appropriate HTTP status codes (200, 401, 400)

Ready to add as comment? Type CONFIRM to post these criteria to the task.

Safety Patterns

⚠️ Confirmation Rules for Comment Posting

When posting acceptance criteria as comments:

  • Always preview: Show comment text before posting
  • Require explicit confirmation: Use "CONFIRM" token before adding comment
  • Verify task: Confirm which task the comment will be added to
  • Review criteria: Ensure criteria are clear and testable before posting

En savoir plus sur les workflows d'écriture sécurisés →

Troubleshooting

Criteria Too Vague

Symptom: Proposed criteria are still unclear

Possible causes:

  • Task description is too vague
  • AI needs more context

Fix: Provide more context: "Make criteria more specific for [domain]. Include [specific requirements]."

Too Many Criteria

Symptom: AI proposes too many acceptance criteria

Possible causes:

  • Task is too complex
  • AI is being overly thorough

Fix: Ask for focus: "Propose only the top 3-5 most important criteria"

Related Tools

This workflow uses these Corcava MCP tools:

get_task

Read task details for analysis

list_task_comments

Check existing context and comments

add_task_comment

Post acceptance criteria as comments

Related Use Cases

Sprint Planning

Add acceptance criteria during sprint planning

Task Handoff

Include acceptance criteria in handoffs

Clarify Task Requirements

Connect your AI assistant to Corcava and turn vague tasks into testable outcomes

Get Started →Tool Reference