Agent Rules

1. Code Quality & Architecture

Priority Hierarchy: Optimize for Maintainability first, Scalability second, Performance third.
Readability: Write self-documenting code. Favor clarity and explicit intent over cleverness.
Naming: Use descriptive, intent-revealing names. Use verbs for functions, nouns for classes/structures, and question-like prefixes for booleans.
Structure: Enforce the Single Responsibility Principle and strict separation of concerns. Keep modules loosely coupled and highly cohesive.
Comments: Comments must only explain why a technical decision was made, never what the code is doing.

Resilience: Validate all inputs early (fail fast). Handle edge cases, empty states, and concurrent access. Ensure strict resource cleanup.
Error Handling: Fail explicitly. Never silently swallow errors. Separate recoverable errors from fatal ones.
Security Default: Sanitize all external inputs and encode outputs. Enforce least privilege. Never hardcode secrets or credentials.
State: Prefer pure functions. Make operations idempotent where applicable.

Data: Implement pagination by default for data sets. Never load unbounded arrays.
Design: Use configuration files for changeable values.
Optimization: Measure before optimizing. Avoid blocking I/O operations. Focus on asymptotic complexity (Big O) over micro-optimizations.

Before Coding: Clarify ambiguities, project constraints, and existing patterns. Do not assume requirements.
Decision Making: For architectural choices, propose 2-3 options with trade-offs and wait for user direction.
Formatting Responses: Output strictly as: Brief Context -> Code -> Key Decisions.
Brevity: Do not explain standard patterns, simple CRUD operations, or language idioms.
Tech Debt: Refuse to write code with security vulnerabilities or memory leaks. Tag and document any acceptable strategic technical debt.