📋 Backlog Grooming

Generate Dependencies and Blocker Analysis

Analyzes a set of upcoming backlog items and produces a dependency graph, blocker list, and sequencing recommendation. Built for PMs planning multi-team initiatives with hidden coupling.

This prompt builds a lightweight dependency map across backlog items using a producer-consumer model, then identifies critical-path blockers and external team dependencies. Output includes a Mermaid-compatible graph definition and a sequencing recommendation.

When to use this prompt

Use this before quarterly planning or when you are lining up the next 2-3 sprints of work across a complex system. It helps expose dependencies that are implicit in your team's heads but not written down anywhere. You will need a short description of each item and, critically, a note on which systems or teams each one touches. The prompt cannot see your code, so it infers dependencies from the language you use; give it specifics (service names, API names) rather than vague feature titles.

The Prompt

Role: Product Manager Variables: {{initiative_name}}, {{backlog_items}}, {{external_teams}}, {{horizon}}

You are a technical product manager analyzing dependencies across a set of upcoming backlog items. Your job is to surface hidden coupling, flag external team dependencies, and propose a sequencing that minimizes blocking.

Initiative: {{initiative_name}}
Items to analyze:
{{backlog_items}}
Known external teams: {{external_teams}}
Time horizon: {{horizon}}

Produce four sections:

1. DEPENDENCY MAP — A Mermaid-compatible graph definition (graph LR) showing which items depend on which. Use the item IDs as nodes. Label edges with the dependency type: data, API, infra, design, or legal.

2. EXTERNAL DEPENDENCIES — A table listing any item that requires work from another team, with columns: item, team, ask, lead time needed, risk if late.

3. CRITICAL PATH — The longest chain of dependent items. Explain why it is the bottleneck.

4. SEQUENCING RECOMMENDATION — The order in which items should be started, with a 1-line rationale each. Explicitly call out any item that should be parallelized.

Assumptions and unknowns: end with a list of things you inferred but could not verify. Ask the user to confirm them before sprint commitment.

Example Output

DEPENDENCY MAP
graph LR
  A[PROD-101 Auth migration] --> B[PROD-102 Profile page]
  A --> C[PROD-103 Settings API]
  C --> D[PROD-104 Notification prefs]
  E[PROD-105 Email template] -.design.-> B

EXTERNAL DEPENDENCIES
| Item | Team | Ask | Lead time | Risk |
| PROD-103 | Platform | New endpoint | 2 weeks | Blocks 3 downstream items |
| PROD-105 | Design | Template file | 1 week | Low, can parallelize |

CRITICAL PATH: PROD-101 to PROD-103 to PROD-104. Three sequential items, all touching auth state. Bottleneck is the auth migration because nothing else can safely ship on the old token format.

SEQUENCING
1. PROD-101 (start immediately, no deps)
2. PROD-105 (parallel, design only)
3. PROD-103 (after 101)
4. PROD-102 (after 101 and 105)

Unknowns to confirm: Platform team capacity for PROD-103, design availability for PROD-105.

Recommended Tools

Jira supports native dependency links and has a Gantt view that visualizes critical paths without a separate tool. Linear has the cleanest UX for blocked-by relationships and will auto-update when upstream items ship. Monday is the right choice when dependencies span multiple boards or departments since its cross-board dependencies are first-class. All three can ingest the Mermaid graph as documentation.

Jira Software75/100 Linear74/100 monday

Frequently Asked Questions

When should I use this prompt?

Run it before quarterly or sprint-over-sprint planning on any initiative touching more than one service or team. It is most valuable for cross-functional initiatives where dependency surprises routinely delay launches. Do not use it for single-team work with tightly scoped changes; the overhead of reading the output exceeds the value. You should also skip it when you already have a current dependency document reviewed in the last 2 weeks.

How accurate is the dependency inference?

The model can only infer dependencies from the language of your item descriptions; it does not know your code. Treat the output as a strawman to walk through with a tech lead rather than a trusted source of truth. In particular, pay attention to the assumptions and unknowns section at the end: any item flagged as uncertain needs a 5-minute conversation with the engineer most familiar with that part of the system before you commit to the sequencing.