product-expert

Deep product analysis through belief networks. Scans issue trackers (GitHub, GitLab, Jira) with a product management lens, extracts factual beliefs about product state, and surfaces risks that dashboards and status meetings miss.

What it finds: Not project delays, but product gaps — features users need that aren't being built, user experience pain points buried across multiple issues, competitive risks from missing table-stakes features, and prioritization misalignment between what's getting built and what matters to users.

How it works: product-expert fetches issues from your tracker, analyzes them through a product lens (user impact, competitive context, product-market fit), extracts factual claims into a reason maintenance system, and synthesizes a product summary from verified beliefs.

Install

uv tool install git+https://github.com/benthomasson/ftl-product-expert

Prerequisites — these CLIs must be on your PATH:

• entry — chronological entry creation
• beliefs or reasons — belief registry management
• claude or gemini — at least one LLM CLI

Platform CLIs (install whichever you need):

• gh — GitHub CLI
• glab — GitLab CLI
• For Jira: set JIRA_URL, JIRA_USER, JIRA_TOKEN env vars

Quick Start

# 1. Point product-expert at an issue tracker
product-expert init github owner/repo --domain "Payment platform"

# 2. Scan issues for a product overview
product-expert scan

# 3. Explore topics one at a time
product-expert explore              # next topic
product-expert explore --pick 3     # specific topic
product-expert explore --pick 1,3,8 # multiple (stable indices)
product-expert explore --skip       # skip and move on

# 4. Extract beliefs from exploration entries
product-expert propose-beliefs
# Review proposed-beliefs.md — mark entries ACCEPT or REJECT
product-expert accept-beliefs

# 5. Build a product summary from beliefs
product-expert summary

# 6. Check progress
product-expert status

How It Works

product-expert follows a scan -> explore -> distill -> reason -> summarize pipeline:

scan              Fetch issues -> product-focused LLM analysis -> topic queue
  |
  v
explore           Pop one topic, fetch full issue, analyze product impact
  |                 |- feature      Feature request deep-dive
  |                 |- epic         Epic with user stories and scope
  |                 |- roadmap      Roadmap alignment assessment
  |                 |- user-story   User story analysis
  |                 |- feedback     User feedback pattern analysis
  |                 +- general      Cross-cutting product analysis
  |
  v
propose-beliefs   Batch-extract factual claims from entries
  |
  v
accept-beliefs    Import reviewed claims into beliefs.md / reasons.db
  |
  v
summary           Synthesize product summary from verified beliefs

Commands

product-expert init <platform> <target>

Bootstrap a knowledge base.

product-expert scan

Fetch issues and produce a product-focused overview covering feature landscape, user impact, product gaps, and prioritization concerns.

product-expert scan --limit 10 --all-pages  # paginate through all issues

product-expert explore

Deep-dive into topics with product analysis: user stories, competitive context, success criteria.

product-expert propose-beliefs / accept-beliefs

Extract and import factual claims about the product.

product-expert summary

Synthesize a product summary from verified beliefs — the document a CPO reads to understand product state in 5 minutes.

product-expert status

Dashboard showing entries, beliefs, topic queue, and cached issues.

Global Options

Option	Description
--model, -m	Model to use: claude or gemini (default: claude)
--quiet, -q	Suppress explanation output to stdout
--timeout, -t	LLM timeout in seconds (default: 300)
--version	Show version