Free download
A problem framing skill for Claude
Product discovery, run by Claude, before anyone builds anything. Give it a ticket, a PRD or a rough idea. It separates what's known from what's only assumed, asks you a few sharp questions at a time, and writes a problem brief for your team and a context file for your AI build tool.
Email only, no name, no company. I'll send you the file and a few notes on doing better product work with AI. Unsubscribe anytime.
What you get back
One brief for people, one for agents
AI has made building cheap, which makes building the wrong thing cheap too. A prototype generated from a fuzzy prompt looks finished and convinces people. These two files are the step before the prompt.
problem-brief.md
The full brief for your team. Problem statement, red flags, the riskiest assumptions with the cheapest test for each, the 5W1H detail with a confidence level on every answer, and open questions with owners.
agent-context.md
A short version for Claude Code, Cursor, Lovable, v0 or Figma Make. Paste it in or save it as a rules file, so the build stays anchored to the problem rather than to whatever the first prompt said.
The framework
5W1H, updated for AI products
Every answer is marked known, assumed or unknown, because teams treat assumptions as facts and that's where the wrong solution comes from. The riskiest assumptions are sorted into value, usability, feasibility and viability risks, each with the cheapest way to test it before building.
Why
What problem this solves and for whom, stated without naming a solution. What evidence says it is real. Whether AI is the answer to the problem or just where the idea started.
What
What the solution must do, what is out of scope, and the constraints. For AI features: what it does on its own, what it only suggests, what it must never do, and what happens when it is wrong.
Who
Users described by their situation and current workaround, not their demographics. The stakeholders who decide, the ones who can stop it, and who signs off on risk.
Where
The device, the other tools that are open, and the environment. Whether the work belongs inside your product or inside the AI tools people already use.
When
The moment the problem shows up and what triggers it, not just the launch date. And how soon something will be in front of real users.
How
One primary success measure, the guardrail that must not get worse, and how you will know the output is good enough before you ship it.
Honest limits
What it isn't
Not user research
It shows you what you don't know. Talking to users is still how you find out.
Not a PRD generator
It frames the problem. Requirements, specs and designs come after, and should be shaped by what the framing found.
Not a decision-maker
It will tell you an idea looks weak and why. Whether to go ahead is still your call.
Questions
What is the 5W1H framework?
A set of six questions (why, what, who, where, when and how) used to explore a problem before solving it. It is often called the Kipling method, after a 1902 poem by Rudyard Kipling. Journalists and quality engineers have used it for over a century, and it works for product discovery for the same reason: it forces you to explore before you create.
What does the skill actually do?
It reads whatever you give it (a ticket, a PRD, a Slack thread or a rough idea), works out what is known and what is only assumed, then asks you up to five questions at a time about the gaps. It checks the request for red flags, such as a solution posing as a problem, ranks the riskiest assumptions and writes a problem brief plus a short context file for your AI build tool.
Who is it for?
Product designers, PMs and engineers about to start a feature, redesign, prototype or vibe-coding session. It is also useful before a kickoff meeting, as a pre-read that highlights the unknowns so the meeting is spent on the gaps.
Does it work in Claude Code, the desktop app and claude.ai?
Yes. Open the file with the Claude desktop app to install it, unzip it into ~/.claude/skills/ for Claude Code, or upload it on claude.ai under Settings, then Capabilities, then Skills.
Is it free?
Yes. You give an email address, the download starts straight away and I send you a copy as a backup, along with a few notes on doing better product work with AI. Unsubscribe anytime.
Get the skill
To install it, open the downloaded file with Claude. Then say something like “help me frame the problem in this ticket” and paste in what you have. On claude.ai, upload it under Settings, then Capabilities, then Skills.
Already framed it and built something? The free UX audit skill covers the other end.