requirement-interviewer
Skill124 starsChanged 5 months agoMITView on GitHub
Gather missing product, UX, technical, and delivery constraints before planning or coding. Use when a request is underspecified, when several implementations are plausible, when frontend work needs explicit acceptance criteria, or when the agent must ask structured questions before acting.
In this repository5 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- frontend-designSkill
- frontend-reviewerSkill
- skill-lifecycleSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 12 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈12
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
Ask only the next unanswered question cluster, not the whole interview at once.
- Very specific: line 24
Read references/answer-completeness.md to decide whether the reply is sufficient or needs follow-up.
- Very specific: line 25
Read references/gating-rules.md before moving from clarification into planning.
- Distinctive: line 26
After each phase, summarize confirmed facts, explicit assumptions, and open questions.
- Distinctive: line 27
Only when all mandatory gates pass, hand off a concise, execution-ready brief or plan.
- Distinctive: line 31
Ask 1-3 short questions at a time.
- Distinctive: line 32
Prefer constrained choices when possible.
- Distinctive: line 33
Keep questions ordered: objective first, implementation later.
- Distinctive: line 34
If the user says "you decide", record the default explicitly as an assumption.
- Distinctive: line 39
Do not generate code, file edits, or architecture proposals before gates pass.
- Distinctive: line 40
Do not ask duplicate questions.
- Distinctive: line 41
Do not jump into low-level implementation details before product intent is clear.
- Distinctive: line 42
Do not silently fill critical gaps with guesses.
- Distinctive: line 43
Stop and ask for explicit approval when the decision changes scope, cost, or stack direction.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Requirement Interviewer
Collect missing constraints through phased questioning. Refuse to plan, code, or propose irreversible implementation details until the required answers are complete.
Workflow
- Read
references/question-flow.md. - Ask only the next unanswered question cluster, not the whole interview at once.
- Read
references/answer-completeness.mdto decide whether the reply is sufficient or needs follow-up. - Read
references/gating-rules.mdbefore moving from clarification into planning. - After each phase, summarize confirmed facts, explicit assumptions, and open questions.
- Only when all mandatory gates pass, hand off a concise, execution-ready brief or plan.
Interview Contract
- Ask 1-3 short questions at a time.
- Prefer constrained choices when possible.
- Keep questions ordered: objective first, implementation later.
- If the user says "you decide", record the default explicitly as an assumption.
- Separate confirmed requirements from inferred assumptions.
Guardrails
- Do not generate code, file edits, or architecture proposals before gates pass.
- Do not ask duplicate questions.
- Do not jump into low-level implementation details before product intent is clear.
- Do not silently fill critical gaps with guesses.
- Stop and ask for explicit approval when the decision changes scope, cost, or stack direction.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

