/AgentReactLoop
Structures a complete autonomous ReAct agent loop: step-by-step thought generation, precise JSON tool invocation schema, and environment observation handling with cycle prevention.
193 commands · 3 subcategories
Structures a complete autonomous ReAct agent loop: step-by-step thought generation, precise JSON tool invocation schema, and environment observation handling with cycle prevention.
Generates an exhaustive RFC-style technical specification covering service architecture, data models, SLAs, latency targets, and failure mode mitigations.
Compares prompt outputs between two model generations (e.g. GPT-4o vs Claude 3.5 Sonnet or Gemini 2.0) across format adherence, tone consistency, and latency.
Write a high-quality prompt for a chain-of-thought reasoning prompt — e.g. a multi-step word problem the model keeps getting wrong.
Harden against manipulation attempts in a persona or character prompt — e.g. a skeptical senior engineer who reviews pull requests.
Build an evaluation harness for a registry of tools an agent can call — e.g. a set of ten internal APIs an agent may invoke.