Task Planning Workflow

The Task Planning workflow retrieves your assigned Azure DevOps work items, enriches them with codebase context, decomposes them into implementation tasks, and organizes the results into structured planning files.

The agent reads your backlog, maps each item to relevant code, generates task breakdowns with priority ordering, and produces a recommended starting point for your next coding session.

When to Use

• 📋 Sprint planning preparation when you need to decompose assigned stories into tasks
• 🏔️ Breaking down epics or features into smaller, actionable work items
• 📅 Organizing your daily work queue with priority-ordered implementation plans
• 🔍 Reviewing assigned items to identify gaps, stale entries, or items needing updates
• 🔄 Refreshing planning files after backlog changes to keep local tracking current

What It Does

1. Retrieves work items assigned to you using mcp_ado_wit_my_work_items with configurable filters
2. Reads planning files from previous sessions to maintain continuity across runs
3. Enriches each work item with codebase context through semantic search for related files
4. Analyzes requirements, acceptance criteria, and existing comments to build implementation context
5. Generates task planning logs with per-item breakdowns, priority ordering, and a recommended top work item
6. Produces a handoff file for downstream workflows (sprint planning, execution, or triage)

NOTE

Task Planning focuses on decomposition and prioritization. It produces planning files consumed by Sprint Planning and Execution workflows. Running task planning before sprint planning gives the sprint planner richer context about each item.

Planning File Structure

The workflow writes its output to a structured directory under .copilot-tracking/workitems/. Each planning session creates or updates files following a consistent template.

The planning-log.md tracks search terms used, work items discovered at each stage, and phase completion status. The work-items.md file serves as the source of truth for planned operations, listing each work item with its action (create, update, or no change), reference number, type, and summary. The handoff.md file packages actions for downstream automation.

Field conventions follow Azure DevOps work item types. User Stories carry title, description, acceptance criteria, story points, and priority. Bugs include repro steps, severity, and priority. Every work item preserves its area path, iteration path, and tags for organizational context.

Work Item Operations

The agent supports creation, updates, batch operations, and linking through MCP ADO tools.

For retrieval, mcp_ado_wit_get_work_item and mcp_ado_wit_get_work_items (batch) hydrate work items with full field values. Batch retrieval reduces API calls when processing multiple items.

The search workflow uses mcp_ado_search_workitem with keyword groups composed from your codebase, commit messages, and existing work item fields. The search protocol supports OR/AND syntax and paging for large result sets.

New work items are created through mcp_ado_wit_create_work_item, and mcp_ado_wit_add_child_work_items decomposes a parent into child tasks. Created items inherit the parent's area path and iteration path unless you override them.

Field changes go through mcp_ado_wit_update_work_item, and mcp_ado_wit_add_work_item_relation handles linking related items. Comments posted through mcp_ado_wit_create_work_item_comment record implementation context directly on the work item.

Output Artifacts

.copilot-tracking/workitems/<planning-type>/<scope-name>/
├── planning-log.md          # Search terms, discovered items, and phase tracking
├── work-items.md            # Source of truth for planned operations
├── task-planning-logs.md    # Per-item breakdowns with priority ordering
└── handoff.md               # Actions packaged for downstream workflows

How to Use

Option 1: Prompt Shortcut

Invoke task planning with a direct prompt:

Process my work items for task planning in my Azure DevOps project

Plan tasks for my assigned items in the current sprint iteration

Option 2: Handoff Button

Click the "Task Planning" handoff button in the ADO Backlog Manager agent to launch with the standard prompt and default configuration.

Option 3: Direct Agent

Start a conversation with the ADO Backlog Manager agent and describe your planning goal. The agent recognizes task planning intent and enters the retrieve-enrich-plan pipeline automatically.

Example Prompts

Full task planning for sprint work:

Process my assigned work items for task planning. Include items in
Active and New states from the current sprint iteration. Enrich each
item with codebase context and recommend:
- Implementation sequence based on dependencies
- Risk areas from related recent changes
- The best starting point for today's work session

Tag-focused planning with priority boost:

Process my assigned work items but focus on items tagged with
"api-redesign." Boost their priority above other items. Limit to
10 items and structure planning files by area path grouping.

Single work item deep planning:

Plan implementation for work item #15432. Pull the full description,
acceptance criteria, and linked items from Azure DevOps. Enrich with
codebase context from the local workspace and generate a detailed
task breakdown.

Output artifacts: Task planning writes planning files to .copilot-tracking/workitems/ containing the prioritized work item list, codebase enrichment results, and recommended start point. Review the enrichment accuracy before feeding output into other workflows.

Tips

• ✅ Run task planning at the start of a sprint to generate fresh planning files from your current assignments
• ✅ Use the boostTags input to prioritize items matching specific tags
• ✅ Review the recommended top work item before starting your coding session
• ✅ Keep planning files current by re-running after backlog changes (reassignments, new items, state updates)
• ❌ Do not skip the enrichment phase (codebase context improves task decomposition accuracy)
• ❌ Do not edit planning files manually if you plan to re-run task planning (the agent overwrites them)
• ❌ Do not run task planning across multiple projects in a single session (scope to one project per run)

Common Pitfalls

Pitfall	Solution
Agent retrieves no work items	Verify your Azure DevOps identity and check that items are assigned to you
Codebase enrichment finds no related files	Ensure the workspace contains the relevant repository (the agent searches local files)
Planning files are stale after backlog changes	Re-run task planning to refresh from current Azure DevOps state
Too many items returned	Use the maxItems input to cap results or filter by iteration path
Handoff file references outdated work item IDs	Clear the planning directory and regenerate from scratch

Next Steps

1. Feed your planning output into the Discovery workflow for work item analysis
2. See Using Workflows Together for the full pipeline walkthrough

TIP

Combine task planning with the forceTopId input when you already know which work item to tackle first. The agent adjusts its recommendation and structures the remaining items accordingly.

---

🤖 Crafted with precision by ✨Copilot following brilliant human instruction, then carefully refined by our team of discerning human reviewers.