| name | office-hours |
| description | Product interrogation before commitment. Challenges whether you're solving the right problem for the right user at the right time. Use when starting a feature, exploring an idea, or before writing a spec. Triggers: "should I build this", "office hours", "is this worth building", "product check", "challenge this idea".
|
| argument-hint | <one-line description of what you're thinking about building> |
Office Hours: Product Interrogation
Challenge whether to build something before committing. Not brainstorming (test
the idea you brought). Not a spec (use /spec-out after). Not council
(product decisions, not technical).
Positioning
- Upstream of council: office-hours challenges whether to build; council
challenges how to build
- Upstream of /spec-out: office-hours challenges the why; spec-out
clarifies the what
- Flow:
/office-hours → /spec-out or /design → council seats
Process
1. Frame the idea
Restate the user's idea in one sentence. Confirm you understand before
questioning.
2. Seven-question framework
Ask questions one at a time. Wait for the answer before asking the next.
Follow up on weak or vague answers before moving on.
Three acts, adaptive questioning:
Act I — The User
Q1. Who specifically?
Name the person, not a persona or segment. "Mobile developers who deploy to
TestFlight weekly" — not "developers." If the user can't name someone specific,
stay on this question.
Q2. What job are they hiring this to do?
A verb and outcome in the user's language, not the builder's. Apply the
Jobs-to-Be-Done framework (see references/jobs-to-be-done.md). The job
exists independent of your solution.
Act II — The Problem
Q3. What happens if you don't build this?
Concrete cost to the user. Not "they'd be disappointed" — what breaks, what
takes longer, what stays impossible? If the answer is "nothing much," that's a
signal.
Q4. What are you assuming about user behavior?
The riskiest assumption, stated as a falsifiable claim. "I assume users will
check this dashboard daily" is testable. "Users want better tools" is not.
Act III — The Bet
Q5. Smallest version that tests the hypothesis?
Not MVP — the experiment. What can you build in days (not weeks) that proves
or disproves the riskiest assumption from Q4?
Q6. What would make you kill this in 3 months?
A measurable kill signal. This question is non-negotiable — always ask it.
If the user can't define failure, they can't define success.
Q7. How much time is this worth?
Shape Up appetite: not how long it will take, but how much you'd bet.
"Two days of effort, max" is an appetite. "However long it takes" is a blank
check.
Adaptation rules
- Minimum 3 questions: Q1, Q3, Q6 (always asked)
- Maximum 7 questions: all seven
- Stay on Act I if the user is poorly defined — don't advance until Q1 has
a real person
- Skip to Q6 if the problem is hypothetical or the user seems to be
exploring rather than committing
- Skip Q2 if the user already described the job clearly in their framing
- Skip Q5 if the user already described a minimal experiment
- Skip Q7 if the user volunteered a time constraint
3. Synthesize the verdict
After questioning, produce the verdict using this template:
## Office Hours: {Title}
**Date:** {YYYY-MM-DD}
**Verdict:** BUILD | NARROW | DEFER | KILL
### The User
{Who, specifically — from Q1}
### The Job
{What they're hiring this to do — from Q2}
### The Hypothesis
{The riskiest assumption, stated as falsifiable claim — from Q4}
### The Bet
{Smallest experiment + appetite — from Q5/Q7}
### Kill Signal
{What kills this in 3 months — from Q6}
### What I Noticed
{Patterns, contradictions, unstated assumptions surfaced during questioning.
Be direct. This is the value-add beyond the user's own answers.}
Verdict criteria
| Verdict | Criteria |
|---|
| BUILD | User real, problem concrete, hypothesis falsifiable, appetite proportional |
| NARROW | Good instincts, scope outpaces evidence — reduce to what you can prove |
| DEFER | Interesting, insufficient evidence — name what to learn first |
| KILL | No real user, hypothetical problem, or unfalsifiable bet — stop here |
Anti-patterns
- Generating ideas: test the idea the user brought, don't brainstorm
alternatives (use
/brainstorm for that)
- Writing specs: the verdict points to next steps, it doesn't define them
- Being encouraging: the value is honest challenge, not validation
- Asking all seven questions mechanically: adapt to what the user reveals