lampson_skill

Overview

This skill encodes Butler Lampson–style system design guidance for architects and engineers. It captures principles of clean abstraction, durable interfaces, capability-based security, and practical hints for functionality, speed, and fault tolerance. Use it to drive architectural decisions, API contracts, and long-lived system design.

How this skill works

The skill inspects design plans and asks Lampson-style review questions: what is the abstraction, its invariants, failure modes, and evolution path. It suggests concrete patterns—capabilities, separation of normal and worst case, write-ahead logging, caching hints—and points out interface leaks and premature optimizations. Recommendations are actionable and prioritized: make it work, then make it right, then make it fast.

When to use it

• Designing public APIs or libraries that must remain stable for years
• Defining security boundaries and permission models (prefer capabilities)
• Architecting storage, replication, and recovery for durability
• Refactoring systems to reduce coupling and hide implementation details
• Preparing design reviews or documenting invariants and failure modes

Best practices

• Define interfaces before implementing; document what the abstraction guarantees and explicitly what it doesn’t
• Keep interfaces minimal, consistent, and hide implementation secrets aggressively
• Optimize only after correctness is established; separate fast normal path from slower worst cases
• Prefer capability-style permissions and principle of least privilege for security
• Design atomic or idempotent operations and use end-to-end verification for correctness

Example use cases

• Design a storage engine interface that hides caching and file layout while guaranteeing atomic puts and durable logs
• Create an API for a distributed service that exposes minimal operations and documents invariants and error modes
• Replace ACL-based access checks with passed capabilities to simplify delegation and reduce confused-deputy risks
• Tune performance by adding a hinting cache that supplies approximate answers and verifies them on demand
• Run a design review checklist to validate evolution plans, failure isolation, and long-term maintainability