What problem does it solve? Non-trivial code changes often get planned from a single perspective, missing trade-offs around simplicity, performance, and risk. This Skill runs three parallel high-reasoning design agents from distinct perspectives, then critically reviews and merges their outputs into one compact, self-contained plan with rejected alternatives documented. ## Core Features & Use Cases - Parallel Multi-Perspective Design: Launches three concurrent agents optimizing for Simplicity & Maintainability, Performance & Scalability, and Minimal Change & Risk. - Strict Read-Only Operation: Audits the working tree with git status and restores a clean state if any agent wrote files, guaranteeing planning never mutates code. - Decision Record Output: Produces a synthesized plan with ordered tasks, acceptance criteria, risks with mitigations, and at least two explicitly rejected alternatives. - Use Case: Before refactoring a data model across a dozen files, run this Skill to get a durable plan comparing three architectural approaches, then hand the result to an execution skill without re-exploring the repo. ## Quick Start Ask the AI to use multi-agent planning to design a plan for migrating the authentication module to token-based sessions, including rejected alternatives.
