audit-backlog-ado

Backlog Audit — Azure DevOps

Purpose

A neglected backlog is a planning liability. Items without acceptance criteria become guesswork in Sprint Planning. Items that haven't moved in months crowd out real priorities. Items sized at 40 points can't be committed to. This audit surfaces all of it before it causes a problem, so the team walks into refinement with a clear picture of what needs attention.

Tool detection

This skill requires ADO MCP tools. If mcp__azure-devops__* tools are not active, tell the user: "This skill requires an Azure DevOps MCP connection. See docs/mcp-setup.md for setup instructions."

Step 1 — Identify the backlog scope

Ask: "Which project and team backlog should I audit? And how deep — just the top 20 items, the full backlog, or a specific number?"

Use wit_list_backlogs to confirm available backlog levels (Portfolio, Requirements, Task).

Default: audit the Requirements backlog (Product Backlog Items and Bugs) — the layer the team commits from. If the team uses a different backlog level, adjust.

Use wit_list_backlog_work_items or wit_query_by_wiql to retrieve the items. Sort by backlog rank (stack rank). Read: title, state, story points, description, acceptance criteria (custom field or description), tags, and last modified date.

Step 2 — Run the health checks

For each backlog item, evaluate the following criteria. Track counts and flag specifics.

2a — Readiness (Definition of Ready)

2b — Size

2c — Staleness

2d — Priority coherence

• Are the top 10 items the ones the PO would actually commit to next Sprint?
• Are there items in the top 10 that depend on items ranked below them? (dependency inversion risk)
• Are there items in Active state that are not in the current Sprint? (orphaned actives)

2e — Sprint assignment anomalies

• Items assigned to a past Sprint that were not completed — are they still relevant or should they be reprioritised?
• Items assigned to a future Sprint with no estimate or AC — they will fail Sprint Planning.

Step 3 — Produce the audit report

Backlog Audit — [PROJECT] / [TEAM] — [DATE]

Items reviewed: [N]
Items flagged: [N] ([X]%)

━━━ Critical — blocks Sprint Planning ━━━━━━━━━━━━━━━━━

Missing acceptance criteria: [N items]
  - #[ID] [Title] — last modified [date]
  - ...

Unestimated items in top 20: [N items]
  - #[ID] [Title]
  - ...

━━━ Size issues ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Oversized items (> [threshold] points): [N]
  - #[ID] [Title] — [N] points — consider splitting

━━━ Staleness ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Items not touched in 90+ days: [N]
  - #[ID] [Title] — last modified [date] — still in [state]

Blocked / On Hold with no path forward: [N]
  - #[ID] [Title] — tagged [tag] since [date]

━━━ Priority concerns ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Dependency inversions detected: [N]
  - #[ID] [Title] ranked above #[ID] [Title] which it depends on

Orphaned actives (Active but not in a Sprint): [N]
  - #[ID] [Title]

━━━ Summary ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Refinement-ready (top [N] items): [N] of [N]
Recommended refinement focus: [list top 3–5 most critical items to fix]

Step 4 — Offer actions

After presenting the report, offer:

1. "Should I add comments to the flagged items explaining what's needed?"

• If yes, use wit_add_work_item_comment for each flagged item with a clear, actionable note: "This item is missing acceptance criteria and cannot be committed to in Sprint Planning. Please add at least one testable AC before the next refinement session."

1. "Should I tag the oversized items so they're easy to filter in the refinement session?"

• If yes, use wit_update_work_item to add a tag like needs-splitting.

1. "Should I generate a refinement agenda — ordered list of items to address in the next session?"

• If yes, produce a prioritised list: critical items first (missing ACs, unestimated), then oversized, then stale.

Never modify state, priority, or content of items without explicit confirmation.

Guardrails

• Never change backlog priority (stack rank) without explicit PO confirmation — rank is a PO decision.
• Never mark items as Resolved or Closed — only flag for PO attention.
• A large story is not automatically bad — if the team consistently delivers 40-point stories because they decompose them into tasks during Sprint Planning, note it but don't flag it as critical.
• If the backlog has more than 100 items, focus the detailed audit on the top 40 (the realistic planning horizon). Note the total backlog size but don't flood the report with items the team won't touch for six months.
• Stale items in the bottom half of the backlog are normal — only flag stale items that are ranked in the top 30.