This skill transforms single-agent ideas into robust, review-validated architectures by simulating a formal peer-review process using multiple constrained AI agents. By assigning specific, non-overlapping roles—such as the Skeptic, Constraint Guardian, and User Advocate—it identifies hidden assumptions, surface failure modes, and validates non-functional requirements before any code is written. It replaces chaotic brainstorming with a gated, sequential workflow that enforces a mandatory Decision Log, ensuring every architectural choice is scrutinized, documented, and approved through a formal arbitration process.
主な機能
01Gated exit criteria to ensure all objections are resolved
02Sequential multi-agent review loop with specialized roles
03Mandatory Decision Log for tracking rationales and alternatives
040 GitHub stars
05Formal arbitration phase to finalize complex technical decisions
06Strict role-based constraints to prevent scope creep
ユースケース
01Conducting automated pre-mortems to identify potential failure modes
02Stress-testing complex system architectures before implementation
03Validating designs against performance, security, and usability constraints
