Universal behavioral principles for AI agents: Karpathy's guidelines adapted for Hermes (SOUL.md) and OpenAI Codex (skill). Generalized four-principle framework — Think Before Acting, Simplicity First, Surgical Changes, Goal-Driven Execution.
Skill Claude CodeCodex
Universal behavioral principles to reduce agent mistakes across ALL task types — coding, research, configuration, file operations, writing, and communication. Derived from Andrej Karpathy's observations on agent pitfalls. Use when the agent needs guidance on thinking before acting, keeping solutions simple, making…
Skill Claude CodeCodex
Behavioral guidelines to reduce agent mistakes across all task types — coding, research, configuration, file operations, and communication. Load for any complex or multi-step task where overcomplication, assumption errors, or scope creep are risks. Derived from Andrej Karpathy's observations on LLM coding pitfalls…