Epic Backlog Analyzer — Claude Plugin

A Claude Code plugin for Technical Product Managers that audits product backlog health before PI planning or quarterly commitment.

What it does

The Epic Backlog Analyzer scores your backlog across four weighted dimensions:

Dimension	Weight	What it checks
Epic Structure Quality	30%	Goal clarity, scope boundaries, outcome metric, definition of done
Story Categorization	25%	Orphan stories, misplaced stories, story-to-epic alignment
Sizing Balance	20%	Epics too large, too small, or unevenly distributed
Theme Uniformity	25%	Jaccard keyword similarity; detects theme drift within epics

Output: per-epic scorecards, an overall health score with band (Healthy / Needs Attention / Significant Issues / Critical), and a prioritized findings report with severity and effort ratings.

Install

Option 1 — Claude Code plugin (recommended)

# Clone the repo
git clone https://github.com/agentbee0/product-agents.git
cd product-agents

# Point Claude Code at the plugin root
# (add to your project's .claude/settings.json)

{
  "plugins": ["path/to/product-agents"]
}

Option 2 — Distribution bundle

# Download and extract the latest bundle, then:
cd dist/po-tpm-skills-bundle
./install.sh

Default install target: ~/.claude/plugins/epic-backlog-analyzer

Usage

As a Claude agent

Once installed, invoke the agent in Claude Code:

"Analyze the backlog in backlog.csv and give me an epic health report."

Claude will parse the CSV, score all four dimensions, and produce a structured findings report.

As a standalone script

python3 skills/epic-backlog-analyzer/scripts/backlog_health_score.py \
  --input backlog.csv \
  --output findings.json

# JSON output only
python3 skills/epic-backlog-analyzer/scripts/backlog_health_score.py \
  --input backlog.csv \
  --json

Required CSV columns: epic, story, epic_goal, epic_dod, epic_scope, epic_outcome, story_description, story_points, persona, domain

Input format

epic,story,epic_goal,epic_dod,epic_scope,epic_outcome,story_description,story_points,persona,domain
User Authentication,Login flow,"Enable secure user access",...,"As a user I can log in",3,end-user,auth

See skills/epic-backlog-analyzer/tests/sample_backlog.csv for a full working example with 4 epics and intentional issues to validate each finding type.

Health score bands

Band	Score	Meaning
✅ Healthy	80 – 100	Backlog is well-structured and ready to plan against
⚠️ Needs Attention	60 – 79	A few structural issues; address before committing
🔶 Significant Issues	40 – 59	Multiple epics have gaps; restructuring recommended
🔴 Critical	0 – 39	Backlog is not ready for planning

Repo structure

product-agents/
├── .claude-plugin/
│   └── plugin.json                  ← Claude Code plugin manifest
├── agents/
│   └── epic-backlog-analyzer.md     ← Agent definition (auto-invocation)
├── skills/
│   └── epic-backlog-analyzer/
│       ├── SKILL.md                 ← Full skill prompt
│       ├── assets/
│       │   ├── backlog-analysis-template.md
│       │   └── generate_carousel.py
│       ├── references/
│       │   └── epic-health-methodology.md
│       ├── scripts/
│       │   └── backlog_health_score.py
│       └── tests/
│           ├── sample_backlog.csv
│           └── sample_findings.json
├── dist/
│   └── po-tpm-skills-bundle/        ← Installable bundle
├── CHANGELOG.md
└── LICENSE

Requirements

Python 3.8+
No external dependencies for backlog_health_score.py
reportlab for carousel PDF generation (pip install reportlab)

License

MIT — see LICENSE.