claude-code-methodology

Design production-grade AI agents by evaluating systems against an 11-dimension framework covering tool design, permissions, orchestration, token economy, memory, and observability. This methodology replaces trial-and-error architecture with structured, production-tested recommendations and safety guardrails.

Review existing multi-agent systems using a structured agent review mode that evaluates token economy, memory compaction, permissions, and task handling. It provides production-grade improvement plans to reduce token spend and harden operational reliability across the 11 dimensions.

The agent design methodology is framework-agnostic, transferring across Python, Go, Rust, LangChain, CrewAI, and custom stacks. It provides best practices for tool design and orchestration without locking you into a single implementation style or framework.

Harden permissions and task lifecycle handling by applying an 11-dimension evaluation framework with safety guardrails. Assess agent systems across permissions, task lifecycle, and conversation flow to generate structured recommendations for secure, production-grade operations.

A structured methodology for AI agent observability and workflow modes is needed when moving agents to production. It evaluates operational tradeoffs across orchestration, extensibility, and observability to ensure reliable conversation flow and task execution.

Multi-agent orchestration and high token usage fail without evaluating token economy, memory management, and orchestration tradeoffs. Apply a production-tested methodology to analyze operational bottlenecks and generate framework-agnostic best practices for agent design.