Corcava logoLe seul outil métier dont vous avez besoinCorcava
Menu

Détection des tâches en double avec MCP : fusionner ou lier le travail similaire

Identify and consolidate duplicate tasks using MCP. This workflow shows you how to find likely duplicate tasks by comparing titles and descriptions, propose merge plans, and safely link related work—with approval steps to prevent accidental changes.

Ce que ce workflow permet

Duplicate task detection with MCP helps you maintain a clean task list:

Résultats clés

  • Duplicate identification: Find tasks with similar titles and descriptions
  • Merge planning: Propose which tasks should be consolidated
  • Safe linking: Add comments or links between related tasks after approval
  • Cleanup recommendations: Get actionable suggestions for task consolidation
  • Approval-based safety: Only add comments/links after explicit confirmation

Prérequis

Before using this workflow, ensure you have:

Workflow pas à pas

Step 1: Find Potential Duplicates

Start by searching for tasks that might be duplicates:

Duplicate Search Prompt

"Search for tasks that might be duplicates: - List all tasks in the project/board - Compare task titles and descriptions - Identify tasks with similar wording or intent - Show me: task pairs that look like duplicates, with their titles and brief descriptions Only analyze, don't make any changes yet."

What the AI does:

  1. Calls list_tasks to get all tasks
  2. Compares task titles and descriptions
  3. Identifies similar or duplicate tasks
  4. Presents potential duplicate pairs

Step 2: Review Task Details

Get full details of potential duplicates to confirm they're the same:

Task Detail Review Prompt

"For the potential duplicate tasks I identified, get full details: - Read each task's description and comments - Compare the actual work being described - Determine if they're true duplicates or just related - Show me: detailed comparison with recommendations This is read-only analysis."

What the AI does:

  1. Calls get_task for each potential duplicate
  2. Reads task descriptions and comments
  3. Compares task content and context
  4. Provides detailed comparison and recommendations

Step 3: Propose Merge Plan

Generate a merge plan for confirmed duplicates:

Merge Plan Prompt

"For the confirmed duplicate tasks, propose a merge plan: - Identify which task should be the primary (keep) - Suggest which task should be merged (close/link) - Explain what information to preserve - Show me: merge plan with task IDs and reasoning Only propose, don't execute any merges."

What the AI does:

  1. Analyzes duplicate pairs
  2. Determines primary vs secondary tasks
  3. Creates merge strategy
  4. Presents merge plan (read-only)

Step 4: Link Tasks (After Approval)

After reviewing the merge plan, link related tasks with comments:

Link Tasks Prompt (Requires Approval)

"Add a comment to task [ID] linking it to task [ID]: - Comment text: 'Related to task #[ID]: [brief description]. Consider merging or consolidating.' - Show me the comment text first for approval - Wait for my confirmation before posting I must type 'CONFIRM' or 'YES, ADD COMMENT' before you add the comment."

What the AI does:

  1. Drafts comment text
  2. Shows preview for approval
  3. Waits for confirmation token
  4. Only calls add_task_comment after approval

Complete Workflow Prompts

Here are complete, copy-paste ready prompts:

Full Duplicate Detection Workflow

"Find and analyze duplicate tasks: 1. Search for tasks that might be duplicates by comparing titles and descriptions 2. Get full details of potential duplicates to confirm they're the same work 3. Propose a merge plan showing which task to keep and which to link/close 4. Show me the complete analysis and recommendations Important: Only analyze and propose. Don't make any changes or add comments unless I explicitly approve with 'CONFIRM' or 'YES, ADD COMMENT'."

Use this when: You want a complete duplicate detection analysis

Quick Duplicate Check

"Find tasks with similar titles in [project/board]. Show me potential duplicates with their task IDs and titles."

Use this when: You just want to quickly identify duplicates

Link Related Tasks (With Approval)

"Link task [ID] to task [ID] by adding a comment. Show me the comment text first, then wait for my 'CONFIRM' before posting."

Use this when: You've identified duplicates and want to link them

Workflow Variants

Variant 1: Exact Duplicate Detection

Find tasks with identical or near-identical titles:

"Find tasks with identical or very similar titles in [project/board]: - Compare titles word-for-word - Flag tasks with 90%+ title similarity - Show me exact duplicates first Focus on exact matches, not just related work."

What happens: AI focuses on title similarity for exact duplicates

Variant 2: Similar Work Detection

Find tasks that describe similar work but with different wording:

"Find tasks that describe similar work in [project/board]: - Compare task descriptions, not just titles - Look for tasks with similar intent or goals - Identify related work that might overlap Show me tasks that are related but may not be exact duplicates."

What happens: AI analyzes descriptions to find similar work

Variant 3: Cross-Project Duplicate Detection

Find duplicates across multiple projects:

"Search for duplicate tasks across all projects: - Compare tasks from different projects - Find work that's been duplicated across projects - Show me: duplicate pairs with project names This helps identify work that should be consolidated across projects."

What happens: AI searches across all projects for duplicates

Example: Complete Duplicate Detection Session

Here's a real example of how a duplicate detection session flows:

User Prompt

"Find duplicate tasks in the 'API Redesign' project. Propose a merge plan."

AI Response (Analysis)

Duplicate Detection Analysis (3 pairs found)

Pair 1: Exact Duplicates

  • Task #123: "Implement user authentication" (created 2026-04-15)
  • Task #145: "Implement user authentication" (created 2026-04-20)
  • Recommendation: Keep #123 (older), link #145 to it

Pair 2: Similar Work

  • Task #156: "Add OAuth 2.0 support" (description: "Implement OAuth for third-party login")
  • Task #167: "OAuth integration" (description: "Add OAuth 2.0 authentication")
  • Recommendation: These describe the same work - merge into #156

Pair 3: Related Work

  • Task #178: "Fix authentication bug"
  • Task #189: "Authentication error handling"
  • Recommendation: Related but different - link them, don't merge

Merge Plan:

  • Link task #145 to #123 with comment: "Duplicate of #123"
  • Link task #167 to #156 with comment: "Same work as #156 - consider merging"
  • Link tasks #178 and #189 as related work

User Approval

"Yes, add the linking comments. I approve."

AI Action (After Approval)

✅ Added comment to task #145: "Related to task #123: Implement user authentication. Consider merging or consolidating."

✅ Added comment to task #167: "Related to task #156: Add OAuth 2.0 support. Consider merging or consolidating."

✅ Added linking comments to tasks #178 and #189

Safety Patterns

✅ Approval-Based Write Operations

This workflow uses explicit approval for any write operations:

  • Read-first: Always analyze before proposing changes
  • Preview comments: Show comment text before posting
  • Confirmation tokens: Require "CONFIRM" or "YES, ADD COMMENT" before adding comments
  • No automatic merges: Only suggests merge plans, never executes them
  • Safe linking: Adds comments to link tasks, doesn't delete or modify tasks

Troubleshooting

No Duplicates Found

Symptom: Analysis returns no duplicate pairs

Possible causes:

  • Tasks are genuinely unique
  • Search scope is too narrow
  • Tasks use very different wording

Fix: Try broader search: "Find similar work, not just exact duplicates" or expand to multiple projects

Too Many False Positives

Symptom: Many tasks flagged as duplicates that aren't

Possible causes:

  • Similar but distinct work
  • Tasks in a series (e.g., "Part 1", "Part 2")

Fix: Ask for stricter matching: "Only flag exact duplicates, not related work"

Comments Not Being Added

Symptom: AI doesn't add comments even after approval

Possible causes:

  • Confirmation token not recognized
  • API key lacks write permissions
  • Task ID incorrect

Fix: Use exact confirmation: "CONFIRM" or "YES, ADD COMMENT". Verify API key has write access.

Related Tools

This workflow uses these Corcava MCP tools:

Related Use Cases

Detect and Consolidate Duplicates

Connect your AI assistant to Corcava and keep your task list clean