product-view

product-view

A Claude Code skill that flips Claude out of code-language and into product-language — describing what the human on the other side of the screen experiences — until you explicitly ask for the code.

Before / after

Without product-view:

When a user signs in, the auth middleware validates the JWT, hits the /me endpoint to fetch the user record, and dispatches a LOGIN_SUCCESS action. The frontend then mounts the dashboard route and the API begins polling for unread notifications.

With product-view:

They tap "Sign In", type their email and password, and a moment later they're looking at their dashboard. A small red dot appears on the bell icon — they have unread messages.

Same flow. The first describes the code. The second describes what actually happened to the person.

What this is

A conversational mode. You flip it on and Claude stays in product-space — what the user sees, what they do, what changes for them, what they feel — until you flip it off or explicitly ask for the code.

Three sub-modes:

Mode	Use when	You get
explain	You want a walkthrough of an existing feature	What the user experiences end-to-end
reframe	You have a bug, ticket, PR, or stack trace	Customer impact in plain terms
plan	You're sketching new functionality	The user's goal and journey, before any implementation

What this isn't

Other skill	What it does	Why this is different
code-to-prd	One-shot scan of a repo → writes PRD .md files	product-view is a live conversational mode, not a document generator
Sachin Rekhi's release-notes skill	Commit URL → user-facing release note	product-view works on any artifact (code, ticket, idea), at any time, not just at ship-time
mattpocock/skills zoom-out	Zooms up within code (modules, callers, glossary)	zoom-out stays inside code; product-view zooms out of code entirely
Multi-perspective code review (e.g. Review Squad)	Multi-subagent panel reviews from several personas	product-view is single-agent, in-line, and the user perspective is the only lens
wondelai/skills (JTBD, lean-startup, etc.)	Teach and apply strategic product frameworks	product-view operates on whatever's in context; no framework prescribed

If you want a generated PRD, a release note, a multi-persona review, or a strategic exercise, those skills exist and they're better at their thing than product-view would be. This skill does one move: it stops the conversation from sneaking code-language into descriptions of the product.

Install

The skill ships as a Claude Code plugin. From inside Claude Code:

/plugin marketplace add tstanmay13/product-view
/plugin install product-view@tstanmay13-product-view

Or, for a manual install, clone into ~/.claude/skills/:

git clone https://github.com/tstanmay13/product-view.git ~/.claude/skills/product-view-repo
ln -s ~/.claude/skills/product-view-repo/skills/product-view ~/.claude/skills/product-view

How to use it

Three activation modes

On-demand — say any of these and the skill kicks in for that turn:

• /product-view
• "from the product perspective"
• "as a user flow"
• "what does this look like to the user / customer / player / reader"
• "translate this"
• "in plain terms"
• "outside-in"

Persistent — turn it on and leave it on:

view on

It stays on for every subsequent message. Turn off with:

view off

Auto-activation — when you ask Claude to explain a feature, walk through how something works, plan new functionality, or summarize a bug ticket, the skill activates on its own. This is intentional: most software explanations are improved by leading with the product view.

It explicitly does not auto-activate on algorithmic, performance, infrastructure, or code-style questions. See When NOT to use it.

Audience flexibility

The "user" isn't always called "user". The skill picks the right word for the context:

• B2B SaaS → customer
• Game → player
• Blog / CMS → reader
• LMS → student
• Health → patient
• Developer product → the developer using your API

If it isn't obvious, the skill asks once.

Why a skill and not just a prompt

You could prompt Claude every turn with "describe this from the user's perspective." That works for one message. The skill adds four things you'd otherwise have to re-prompt for, every time: