[Release] anthropics/claude-code v2.1.91: v2.1.91

Claude Code v2.1.91 introduces MCP tool result persistence overrides supporting up to 500K characters, adds shell execution disabling options, enables multi-line prompts in deep links, and allows plugins to ship executables. The release includes critical fixes for transcript chain breaks, terminal key handling, plan mode persistence in remote sessions, and JSON schema validation, plus performance improvements and enhanced API guidance.

Key Points

• •MCP tool results can now persist up to 500K characters via `_meta["anthropic/maxResultSizeChars"]` annotation, enabling larger payloads like database schemas without truncation
• •New `disableSkillShellExecution` setting prevents inline shell execution in skills, custom slash commands, and plugin commands for enhanced security control
• •Multi-line prompts now supported in `claude-cli://open?q=` deep links with encoded newlines (`%0A`), improving prompt flexibility
• •Plugins can ship executables in `bin/` directory and invoke them as bare commands from the Bash tool, expanding plugin capabilities
• •Fixed critical transcript chain breaks on `--resume` that could silently lose conversation history due to async write failures
• •Resolved terminal compatibility issues with `cmd+delete` key binding across iTerm2, kitty, WezTerm, Ghostty, and Windows Terminal
• •Fixed plan mode in remote sessions losing track after container restart, preventing permission prompts and empty approval modals
• •Corrected JSON schema validation for `permissions.defaultMode: "auto"` in settings.json configuration
• •Improved `/claude-api` skill guidance with agent design patterns covering tool surface decisions, context management, and caching strategies
• •Performance optimization: faster `stripAnsi` on Bun via native `Bun.stripANSI` routing; Edit tool uses shorter anchors to reduce output tokens

Workflow Diagram

Concepts