golden-rules

What problem does it solve? Non-technical users often give raw, ambiguous requests, and AI assistants tend to guess, over-deliver, or add unrequested changes that break things. This Skill forces a disciplined workflow: reformulate the intent, confirm it, propose a plan, get approval, and only then write code. ## Core Features & Use Cases - Intent Reframing: Restates the user's raw request in one or two sentences and asks a single targeted question when something load-bearing is ambiguous. - Plan-Before-Code Gate: Requires a short plan listing steps, affected files, and a success criterion, validated by the user before any code is written. - Five Execution Rules: Think before writing, keep it simple, one thing at a time, surgical changes only, and exact fidelity to the request with no unrequested extras. - Use Case: A founder says "make the login better." Instead of rewriting auth, the assistant reframes the goal, asks what "better" means, proposes a minimal plan touching only the relevant files, and waits for approval. ## Quick Start Invoke the golden-rules skill at the start of a session or whenever you give a rough idea, so the assistant reframes your intent and validates a plan before coding.