Claude Code 2.1: The Pain Points? Fixed (paddo.dev)

Claude Code 2.1 was released today, introducing a substantial update aimed at addressing user frustrations from previous versions. This release features 109 changes, most notably enhancing skill controllability, eliminating the unfettered auto-activation of skills, and integrating explicit invocation options via a slash command menu. Users can now create and modify skills without restarting the system, see ongoing progress during execution, and run skills in isolated contexts to prevent interference in their main conversations. These changes aim to streamline user experience, reduce reliance on semantic matching for skill activation, and provide clearer visibility into skill operations. The update also introduces hooks as a fundamental aspect of the system, evolving them into first-class citizens for defining guardrails across agents, skills, and commands. This shift empowers developers to enforce deterministic behavior easily, enhancing the robustness of Claude’s execution. Additionally, improvements in long-running operations and the self-verification capabilities of Claude contribute to a more reliable and efficient user experience. While some limitations remain, such as the absence of an explicit invocation flag and invocation logging, the enhancements position Claude Code 2.1 as a more user-centric tool, emphasizing control and reliability in AI-assisted coding.