// Quick user-centric interview to capture requirements from a time-poor stakeholder.
Commit, run bin/ci, and open a PR if CI passes. Use when ready to ship changes.
Convert feature requirements into structured end-to-end test cases as JSON.
Interview me about the plan
Catch up on codebase changes by reviewing the diff from master and reading relevant spec files.
Prime yourself on a codebase by listing files and reading the README.
Reset the working directory for the next task by ensuring no uncommitted changes exist, then switching to the main branch and pulling latest.
| name | discovery |
| description | Quick user-centric interview to capture requirements from a time-poor stakeholder. |
| argument-hint | ["topic or context"] |
| model | opus |
| disable-model-invocation | true |
You are conducting a quick user discovery interview. The user is time-poor (on Slack or a phone call), so you need to capture the essentials efficiently - not 2 questions, not 200, but around 5-10 focused questions that get to the heart of what they need.
The user has provided context: $1
Use AskUserQuestion to ask focused, punchy questions one at a time. Cover these areas (but adapt based on responses):
Don't ask all of these robotically - listen to their answers and follow up where needed. Skip questions that have already been answered. Respect their time.
When the interview is complete, generate a filename using: DISCOVERY-YYYY-MM-DD-<short-summary>.md where <short-summary> is 2-4 lowercase words from the topic (use bash date command to get the date).
Write a concise discovery document:
# Discovery: <Topic>
**Date:** YYYY-MM-DD
**Stakeholder:** [if mentioned]
## User Context
- Who: ...
- Role/situation: ...
## Problem
- Current workflow: ...
- Pain points: ...
## Desired Outcome
- What success looks like: ...
- Frequency/urgency: ...
## Constraints
- Must-haves: ...
- Blockers: ...
## Raw Notes
- [Key quotes or details captured during interview]
Keep it scannable. This doc can feed into /interview for technical deep-dive later.