Interview me about a plan to flesh out requirements before implementation. Asks non-obvious questions one at a time until all ambiguities are resolved.
Systematically interview the user about their plan/PRD to surface and resolve ambiguities BEFORE implementation begins. This prevents wasted iterations in Ralph loops by ensuring the spec is tight.
Credit: @trq212's interview method, documented by @callam53.
When invoked with /interview <plan-file>:
Systematically cover these areas:
Always use AskUserQuestion with a single, focused question. Never batch questions.
Bad: "What technology should we use, and what about error handling, and also what's the timeline?"
Good: "For the transcription service - Recall.ai vs Google Speech API vs Deepgram. Which direction are you leaning and why?"
When an answer reveals new ambiguity, follow up immediately:
User: "We'll use Recall.ai for transcription"
Follow-up: "Recall.ai has both real-time webhooks and polling. For your latency needs, which pattern makes more sense?"
Skip questions with obvious answers. Focus on:
Stop interviewing when:
When interview is complete:
Summarize findings in structured format:
## Interview Summary: [Plan Name]
### Decisions Made
- [Decision 1]: [Choice and rationale]
- [Decision 2]: [Choice and rationale]
### Clarifications
- [Ambiguity]: [Resolution]
### Edge Cases Addressed
- [Case]: [How to handle]
### Remaining Open Questions
- [Question]: [Why it's ok to defer]
### Constraints Confirmed
- [Constraint]: [Details]
Ask where to write the summary:
plan-interview-notes.md)Update the plan file if user wants inline updates
/interview docs/atlas-on-calls/PROMPT.md
[Reads file]
"I've read the Atlas on Calls plan. I have questions about several areas. Let's start with the most critical:
For trigger detection - you mention 'atlas' as the wake word. What should happen if someone says 'Atlas is our AI assistant' (mention but not a question) vs 'Atlas, what's our revenue?' (actual query). How sophisticated does this detection need to be for v1?"
[User answers]
"Got it - simple pattern matching for v1, upgrade later. Follow-up: For the chat trigger '@atlas', should it require the @ symbol, or also match 'atlas,' or 'hey atlas'?"
[User answers]
"Clear. Moving to a different area - Recall.ai pricing. Have you confirmed their pricing model works for your expected call volume? This could affect architecture if there are per-minute costs."
[... continues until done ...]
"I think we've covered the critical ambiguities. Here's a summary of what we resolved:
[Summary]
Where should I save these notes?"
This skill fits in the pre-Ralph workflow:
1. Write rough prd.md / plan idea
2. /brainstorming (optional - if approach unclear)
3. /interview prd.md <-- THIS SKILL
4. Update plan with interview findings
5. /ralph-loop or ralph-import
The interview ensures your Ralph prompt is tight enough that Claude doesn't waste iterations on ambiguous requirements.
During interview:
At summary:
Search for places (restaurants, cafes, etc.) via Google Places API proxy on localhost.
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.
Start voice calls via the OpenClaw voice-call plugin.
Notion API for creating and managing pages, databases, and blocks.
Gemini CLI for one-shot Q&A, summaries, and generation.
Category:developer