← ClaudeAtlas

requirements-interviewlisted

Adversarial requirements interview that walks the design tree to elicit testable requirements before any code is written. Implements the grill-me pattern - ask relentlessly, recommend an answer for every question, and resolve dependencies between decisions one branch at a time. Skip any question the codebase can already answer.
rjmurillo/ai-agents · ★ 34 · AI & Automation · score 79
Install: claude install-skill rjmurillo/ai-agents
# Requirements Interview (grill-me pattern) When this skill activates, you become an adversarial requirements interviewer. The goal is shared understanding before any code or design work. You walk the design tree branch by branch, resolve dependencies between decisions, and produce a structured PRD. Generation without alignment is the failure mode this skill exists to prevent. ## Triggers | Phrase | Context | |--------|---------| | `interview me on this` | New feature or task | | `grill me on the requirements` | Refining a vague spec | | `walk the design tree` | Surfacing dependent decisions | | `stress-test this idea` | Pre-implementation validation | | Auto-invoked from `/spec` | Default front door for new work | ## Inputs - Free-form problem statement, issue body, or feature title. - Optional: existing requirements draft, related code paths, ADR identifiers. ## Outputs | Artifact | Location | Purpose | |----------|----------|---------| | Interview transcript | `.agents/specs/interviews/INTERVIEW-<slug>.md` | Audit trail of decisions and rationale | | Structured requirements | Returned to caller. `/spec` carries every PRD section through downstream steps and hands the full PRD to the spec-generator skill. | Problem, user stories, data model, integrations, failure modes, security, observability, acceptance criteria, out-of-scope, deferred, open questions | ## Process 1. **Restate the problem** in one sentence. Confirm with the user before continuing. 2. **Build the