Claude Code 2.1.261 adds skill diagnostics and larger agent-output controls

Anthropic has released Claude Code 2.1.261, a maintenance-heavy update with several changes that matter for teams running longer agentic coding sessions.\n\nThe most visible addition is /skill-doctor. It reports which loaded skills go unused and how much context they consume, giving users a direct way to identify skill definitions that add prompt overhead without contributing to the session. Claude Code also adds bashOutputMaxChars and taskOutputMaxChars, which can raise the amount of command and background-task output kept inline for Claude to read to as much as 128K characters before the remainder is saved to a file.\n\nFor managed environments, /status and claude doctor now expose an Organization policy line that explains why policy loading failed, including cases where a proxy does not pass the required endpoint through. The release also adds a file-based option for oversized subagent system prompts.\n\nThe rest of the release is dominated by reliability and safety fixes: Remote Control state and stop handling, cloud/session resume context, TLS-inspecting proxy behavior, stuck background-agent CPU loops, connector retries and additional dangerous-rm detection all receive changes. These are concrete product-behavior updates, not independent evidence that Claude Code performs better than competing coding agents.\n\nFor AiToolMap, this warrants a material review refresh because the current Claude Code review predates these context-management and managed-environment controls. The review should revalidate capabilities and operational caveats while leaving its score unchanged unless the evidence base supports a rating change.