Show HN: A Claude Code skill that tells Claude when not to use Claude (github.com)

A newly developed Agent Skill, known as "Claude Code," provides users with a powerful tool to audit codebases for unnecessary calls to large language models (LLMs), such as those from OpenAI and Anthropic. This skill allows developers to identify which LLM API calls can be replaced with deterministic logic, helping to optimize performance and reduce costs without altering the original code. By simply integrating the skill into their agents, users can ask it to assess LLM usage, classify tasks, and generate a detailed Markdown audit report, including cost estimates and recommendations for deterministic alternatives. This development is significant for the AI/ML community as it addresses concerns around the efficiency and cost-effectiveness of LLM deployment. By categorizing API calls into three verdicts—green (🟢), yellow (🟡), and red (🔴)—the skill enables developers to understand the potential for replacements and the associated risks. The focus on transparency and user control, where automatic code alteration is not included by default, encourages a thoughtful approach to employing AI in software development, potentially paving the way for more sustainable and efficient AI applications across various projects.